{
    "说明": "如果中英文环境不同，用models.ChatGPT.desc的写法，如果中英文相同，可以直接用字符串",
    "description": {
        "zh-CN": "OpenCompass 学术开源评测集目前聚焦大语言模型综合推理能力。\n- 目前选取的数据集包括 综合知识推理 (MMLU-Pro /GPQA-Diamond(repeat-4) /HLE) ,数学推理 (AIME2025(repeat-32)) ,代码生成 (LiveCodeBench(repeat-6)) ,指令跟随 (IFEval)。\n- 评测集会以月度为单位进行持续更新，纳入社区最新的研究进展。\n- 当前模型评测以对话模型为主，不定期更新社区最新模型，相关提示词和复现脚本参考OpenCompass代码仓库。\n- 模型评测结果会随着工具链的迭代同步更新。\n- OpenCompass项目中提供了复现学术榜单的配置文件，具体请查阅官方文档。",
        "en-US": "The CompassAcademic currently focuses on the comprehensive reasoning abilities of LLMs.\n- The datasets selected so far include General Knowledge Reasoning (MMLU-Pro /GPQA-Diamond(repeat-4) /HLE), Mathematical Reasoning (AIME2025(repeat-32)), Code Completion (LiveCodeBench(repeat-6)), and Instruction Following (IFEval).\n- The benchmark is updated monthly to incorporate the latest research advancements from the community.\n- Currently, the evaluation primarily targets chat models, with updates featuring the latest community models at irregular intervals. Prompts and reproduction scripts can be found in the OpenCompass toolkit.\n- The model evaluation results will be updated synchronously with the iteration of the toolchain. \n- OpenCompass provides configurations for reproducing this leaderboard. Please refer to the documentation for details."
    },
    "OverallTable": [
        {
            "key": 0,
            "model": "MiniCPM-SALA",
            "org": "OpenBMB",
            "num": "9B",
            "time": "2026/2/11",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 58.7,
            "HLE": 9.3,
            "AIME2025": 81.0,
            "MMLU-Pro": 72.0,
            "LiveCodeBenchV6": 54.7,
            "GPQA-Diamond": 58.2,
            "IFEval": 76.9
        },
        {
            "key": 1,
            "model": "Qwen3.5-397B-A17B",
            "org": "Qwen",
            "num": "397B",
            "time": "2026/2/16",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 78.4,
            "HLE": 27.5,
            "AIME2025": 92.3,
            "MMLU-Pro": 87.6,
            "LiveCodeBenchV6": 83.0,
            "GPQA-Diamond": 88.4,
            "IFEval": 91.5
        },
        {
            "key": 2,
            "model": "GLM-5",
            "org": "Zhipu AI",
            "num": "744B",
            "time": "2026/2/12",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 79.0,
            "HLE": 28.1,
            "AIME2025": 95.8,
            "MMLU-Pro": 85.2,
            "LiveCodeBenchV6": 86.2,
            "GPQA-Diamond": 85.3,
            "IFEval": 93.2
        },
        {
            "key": 3,
            "model": "MiniMax-M2.5",
            "org": "MiniMax",
            "num": "230B",
            "time": "2026/2/12",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 73.3,
            "HLE": 22.2,
            "AIME2025": 86.2,
            "MMLU-Pro": 81.7,
            "LiveCodeBenchV6": 73.6,
            "GPQA-Diamond": 84.6,
            "IFEval": 91.1
        },
        {
            "key": 4,
            "model": "Kimi-K2.5",
            "org": "Moonshot",
            "num": "1000B",
            "time": "2026/1/27",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 78.2,
            "HLE": 28.6,
            "AIME2025": 91.9,
            "MMLU-Pro": 86.2,
            "LiveCodeBenchV6": 80.6,
            "GPQA-Diamond": 88.1,
            "IFEval": 93.9
        },
        {
            "key": 5,
            "model": "Step-3.5-Flash",
            "org": "StepFun",
            "num": "196B",
            "time": "2026/2/2",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 76.9,
            "HLE": 21.6,
            "AIME2025": 95.7,
            "MMLU-Pro": 83.5,
            "LiveCodeBenchV6": 83.9,
            "GPQA-Diamond": 83.7,
            "IFEval": 93.2
        },
        {
            "key": 6,
            "model": "TeleChat-thinking",
            "org": "TeleAI",
            "num": "N/A",
            "time": "2025/12/29",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 71.8,
            "HLE": 20.3,
            "AIME2025": 79.0,
            "MMLU-Pro": 84.1,
            "LiveCodeBenchV6": 71.7,
            "GPQA-Diamond": 81.4,
            "IFEval": 94.3
        },
        {
            "key": 7,
            "model": "GLM-4.7",
            "org": "Zhipu AI",
            "num": "355B",
            "time": "2025/12/23",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 77.6,
            "HLE": 25.4,
            "AIME2025": 95.4,
            "MMLU-Pro": 84.0,
            "LiveCodeBenchV6": 83.8,
            "GPQA-Diamond": 86.9,
            "IFEval": 90.2
        },
        {
            "key": 8,
            "model": "MiMo-V2-Flash",
            "org": "Xiaomi",
            "num": "309B",
            "time": "2025/12/17",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 73.3,
            "HLE": 20.5,
            "AIME2025": 92.9,
            "MMLU-Pro": 83.1,
            "LiveCodeBenchV6": 71.8,
            "GPQA-Diamond": 82.1,
            "IFEval": 89.5
        },
        {
            "key": 9,
            "model": "DeepSeek-V3.2",
            "org": "DeepSeek",
            "num": "671B",
            "time": "2025/12/1",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 75.3,
            "HLE": 23.2,
            "AIME2025": 93.0,
            "MMLU-Pro": 85.8,
            "LiveCodeBenchV6": 75.4,
            "GPQA-Diamond": 84.6,
            "IFEval": 89.7
        },
        {
            "key": 10,
            "model": "DeepSeek-V3.2-Speciale",
            "org": "DeepSeek",
            "num": "671B",
            "time": "2025/12/1",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 78.2,
            "HLE": 28.6,
            "AIME2025": 96.0,
            "MMLU-Pro": 85.5,
            "LiveCodeBenchV6": 80.9,
            "GPQA-Diamond": 86.7,
            "IFEval": 91.7
        },
        {
            "key": 11,
            "model": "Gemini-3-Pro-Preview",
            "org": "Google",
            "num": "N/A",
            "time": "2025/11/19",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 81.3,
            "HLE": 38.0,
            "AIME2025": 93.4,
            "MMLU-Pro": 89.3,
            "LiveCodeBenchV6": 82.9,
            "GPQA-Diamond": 91.5,
            "IFEval": 92.8
        },
        {
            "key": 12,
            "model": "Kimi-K2-Thinking",
            "org": "Moonshot",
            "num": "1000B",
            "time": "2025/11/6",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 75.3,
            "HLE": 21.3,
            "AIME2025": 94.1,
            "MMLU-Pro": 84.3,
            "LiveCodeBenchV6": 77.1,
            "GPQA-Diamond": 82.7,
            "IFEval": 92.4
        },
        {
            "key": 13,
            "model": "MiniMax-M2",
            "org": "MiniMax",
            "num": "230B",
            "time": "2025/10/25",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 69.5,
            "HLE": 13.4,
            "AIME2025": 79.1,
            "MMLU-Pro": 81.6,
            "LiveCodeBenchV6": 74.0,
            "GPQA-Diamond": 78.7,
            "IFEval": 90.2
        },
        {
            "key": 14,
            "model": "JT-Think",
            "org": "China Mobile",
            "num": "N/A",
            "time": "2025/10/22",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 67.2,
            "HLE": 14.4,
            "AIME2025": 65.9,
            "MMLU-Pro": 85.6,
            "LiveCodeBenchV6": 64.9,
            "GPQA-Diamond": 77.7,
            "IFEval": 94.6
        },
        {
            "key": 15,
            "model": "GLM-4.6",
            "org": "Zhipu AI",
            "num": "355B",
            "time": "2025/9/30",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 73.3,
            "HLE": 19.3,
            "AIME2025": 90.3,
            "MMLU-Pro": 83.0,
            "LiveCodeBenchV6": 78.2,
            "GPQA-Diamond": 80.4,
            "IFEval": 88.7
        },
        {
            "key": 16,
            "model": "Qwen3-Next-80B-A3B-Thinking",
            "org": "Alibaba",
            "num": "80B",
            "time": "2025/9/12",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 69.5,
            "HLE": 13.5,
            "AIME2025": 89.0,
            "MMLU-Pro": 82.0,
            "LiveCodeBenchV6": 66.3,
            "GPQA-Diamond": 77.0,
            "IFEval": 89.5
        },
        {
            "key": 17,
            "model": "Qwen3-Next-80B-A3B-Instruct",
            "org": "Alibaba",
            "num": "80B",
            "time": "2025/9/12",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 62.5,
            "HLE": 8.0,
            "AIME2025": 69.2,
            "MMLU-Pro": 81.3,
            "LiveCodeBenchV6": 54.8,
            "GPQA-Diamond": 74.1,
            "IFEval": 87.6
        },
        {
            "key": 18,
            "model": "ERNIE-4.5-21B-A3B-Thinking",
            "org": "Baidu",
            "num": "21B",
            "time": "2025/9/9",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 57.0,
            "HLE": 6.5,
            "AIME2025": 76.2,
            "MMLU-Pro": 70.8,
            "LiveCodeBenchV6": 47.0,
            "GPQA-Diamond": 60.2,
            "IFEval": 81.2
        },
        {
            "key": 19,
            "model": "LongCat-Flash-Chat",
            "org": "Meituan",
            "num": "562B",
            "time": "2025/9/1",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 59.8,
            "HLE": 8.5,
            "AIME2025": 61.0,
            "MMLU-Pro": 81.0,
            "LiveCodeBenchV6": 48.2,
            "GPQA-Diamond": 70.1,
            "IFEval": 90.2
        },
        {
            "key": 20,
            "model": "Seed-OSS-36B-Instruct",
            "org": "ByteDance",
            "num": "36B",
            "time": "2025/8/20",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 66.5,
            "HLE": 9.5,
            "AIME2025": 86.0,
            "MMLU-Pro": 80.6,
            "LiveCodeBenchV6": 67.3,
            "GPQA-Diamond": 71.5,
            "IFEval": 83.7
        },
        {
            "key": 21,
            "model": "Qwen3-4B-Thinking-2507",
            "org": "Alibaba",
            "num": "4B",
            "time": "2025/8/7",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 60.6,
            "HLE": 6.0,
            "AIME2025": 80.0,
            "MMLU-Pro": 72.8,
            "LiveCodeBenchV6": 51.6,
            "GPQA-Diamond": 64.7,
            "IFEval": 88.5
        },
        {
            "key": 22,
            "model": "Qwen3-4B-Instruct-2507",
            "org": "Alibaba",
            "num": "4B",
            "time": "2025/8/7",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 47.2,
            "HLE": 5.1,
            "AIME2025": 46.9,
            "MMLU-Pro": 63.0,
            "LiveCodeBenchV6": 33.5,
            "GPQA-Diamond": 52.3,
            "IFEval": 82.4
        },
        {
            "key": 23,
            "model": "GPT-5-2025-08-07 (high)",
            "org": "OpenAI",
            "num": "N/A",
            "time": "2025/8/7",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 78.8,
            "HLE": 28.5,
            "AIME2025": 94.1,
            "MMLU-Pro": 86.2,
            "LiveCodeBenchV6": 83.2,
            "GPQA-Diamond": 86.4,
            "IFEval": 94.6
        },
        {
            "key": 24,
            "model": "GPT-5-mini-2025-08-07 (high)",
            "org": "OpenAI",
            "num": "N/A",
            "time": "2025/8/7",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 73.2,
            "HLE": 21.5,
            "AIME2025": 90.9,
            "MMLU-Pro": 70.2,
            "LiveCodeBenchV6": 80.8,
            "GPQA-Diamond": 82.7,
            "IFEval": 92.8
        },
        {
            "key": 25,
            "model": "GPT-5-nano-2025-08-07 (high)",
            "org": "OpenAI",
            "num": "N/A",
            "time": "2025/8/7",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 67.2,
            "HLE": 11.0,
            "AIME2025": 85.4,
            "MMLU-Pro": 75.0,
            "LiveCodeBenchV6": 74.0,
            "GPQA-Diamond": 67.8,
            "IFEval": 90.2
        },
        {
            "key": 26,
            "model": "gpt-oss-120b (high)",
            "org": "OpenAI",
            "num": "120B",
            "time": "2025/8/5",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 73.2,
            "HLE": 18.3,
            "AIME2025": 93.4,
            "MMLU-Pro": 79.7,
            "LiveCodeBenchV6": 78.4,
            "GPQA-Diamond": 78.9,
            "IFEval": 90.2
        },
        {
            "key": 27,
            "model": "gpt-oss-20b (high)",
            "org": "OpenAI",
            "num": "20B",
            "time": "2025/8/5",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 66.4,
            "HLE": 11.6,
            "AIME2025": 87.9,
            "MMLU-Pro": 72.8,
            "LiveCodeBenchV6": 68.4,
            "GPQA-Diamond": 68.9,
            "IFEval": 88.9
        },
        {
            "key": 28,
            "model": "Qwen3-30B-A3B-Thinking-2507",
            "org": "Alibaba",
            "num": "30B",
            "time": "2025/7/30",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 66.2,
            "HLE": 11.7,
            "AIME2025": 86.8,
            "MMLU-Pro": 79.5,
            "LiveCodeBenchV6": 59.4,
            "GPQA-Diamond": 70.1,
            "IFEval": 89.7
        },
        {
            "key": 29,
            "model": "Qwen3-30B-A3B-Instruct-2507",
            "org": "Alibaba",
            "num": "30B",
            "time": "2025/7/30",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 55.3,
            "HLE": 7.7,
            "AIME2025": 63.8,
            "MMLU-Pro": 73.9,
            "LiveCodeBenchV6": 38.7,
            "GPQA-Diamond": 63.6,
            "IFEval": 83.9
        },
        {
            "key": 30,
            "model": "GLM-4.5",
            "org": "Zhipu AI",
            "num": "358B",
            "time": "2025/7/29",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 69.2,
            "HLE": 16.9,
            "AIME2025": 85.8,
            "MMLU-Pro": 82.7,
            "LiveCodeBenchV6": 65.0,
            "GPQA-Diamond": 79.5,
            "IFEval": 85.4
        },
        {
            "key": 31,
            "model": "Qwen3-235B-A22B-Thinking-2507",
            "org": "Alibaba",
            "num": "235B",
            "time": "2025/7/25",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 71.8,
            "HLE": 18.5,
            "AIME2025": 90.9,
            "MMLU-Pro": 83.5,
            "LiveCodeBenchV6": 70.6,
            "GPQA-Diamond": 79.8,
            "IFEval": 87.8
        },
        {
            "key": 32,
            "model": "Qwen3-235B-A22B-Instruct-2507",
            "org": "Alibaba",
            "num": "235B",
            "time": "2025/7/22",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 61.3,
            "HLE": 12.3,
            "AIME2025": 69.5,
            "MMLU-Pro": 79.2,
            "LiveCodeBenchV6": 43.0,
            "GPQA-Diamond": 75.5,
            "IFEval": 88.3
        },
        {
            "key": 33,
            "model": "Kimi-K2-Instruct",
            "org": "Moonshot",
            "num": "1000B",
            "time": "2025/7/11",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 57.0,
            "HLE": 6.7,
            "AIME2025": 49.1,
            "MMLU-Pro": 78.4,
            "LiveCodeBenchV6": 46.8,
            "GPQA-Diamond": 72.7,
            "IFEval": 88.5
        },
        {
            "key": 34,
            "model": "ERNIE-4.5-300B-A47B-PT",
            "org": "Baidu",
            "num": "300B",
            "time": "2025/6/30",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 54.7,
            "HLE": 3.8,
            "AIME2025": 32.6,
            "MMLU-Pro": 78.7,
            "LiveCodeBenchV6": 41.0,
            "GPQA-Diamond": 83.0,
            "IFEval": 88.9
        },
        {
            "key": 35,
            "model": "Hunyuan-A13B-Instruct",
            "org": "Tencent",
            "num": "80B",
            "time": "2025/6/27",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 52.3,
            "HLE": 6.0,
            "AIME2025": 65.7,
            "MMLU-Pro": 73.8,
            "LiveCodeBenchV6": 43.9,
            "GPQA-Diamond": 64.1,
            "IFEval": 60.3
        },
        {
            "key": 36,
            "model": "Finix-P1-32B (Thinking)",
            "org": "Ant Group",
            "num": "N/A",
            "time": "2025/6/24",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 62.0,
            "HLE": 6.9,
            "AIME2025": 70.2,
            "MMLU-Pro": 82.7,
            "LiveCodeBenchV6": 54.9,
            "GPQA-Diamond": 68.7,
            "IFEval": 88.7
        },
        {
            "key": 37,
            "model": "Gemini-2.5-Pro",
            "org": "Google",
            "num": "N/A",
            "time": "2025/6/18",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 73.6,
            "HLE": 21.1,
            "AIME2025": 88.7,
            "MMLU-Pro": 85.8,
            "LiveCodeBenchV6": 71.3,
            "GPQA-Diamond": 84.7,
            "IFEval": 90.0
        },
        {
            "key": 38,
            "model": "doubao-seed-1.6-thinking-250615",
            "org": "ByteDance",
            "num": "N/A",
            "time": "2025/6/11",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 66.0,
            "HLE": 13.5,
            "AIME2025": 82.8,
            "MMLU-Pro": 83.1,
            "LiveCodeBenchV6": 61.8,
            "GPQA-Diamond": 75.9,
            "IFEval": 78.9
        },
        {
            "key": 39,
            "model": "DeepSeek-R1-0528",
            "org": "DeepSeek",
            "num": "671B",
            "time": "2025/5/28",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 68.1,
            "HLE": 14.4,
            "AIME2025": 89.0,
            "MMLU-Pro": 83.5,
            "LiveCodeBenchV6": 61.0,
            "GPQA-Diamond": 80.6,
            "IFEval": 80.0
        },
        {
            "key": 40,
            "model": "Claude Sonnet 4 (Thinking)",
            "org": "Anthropic",
            "num": "N/A",
            "time": "2025/5/22",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 61.8,
            "HLE": 8.7,
            "AIME2025": 68.7,
            "MMLU-Pro": 83.0,
            "LiveCodeBenchV6": 47.5,
            "GPQA-Diamond": 74.6,
            "IFEval": 88.3
        },
        {
            "key": 41,
            "model": "Qwen3-32B (Thinking)",
            "org": "Alibaba",
            "num": "32B",
            "time": "2025/4/29",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 61.3,
            "HLE": 8.5,
            "AIME2025": 70.3,
            "MMLU-Pro": 78.0,
            "LiveCodeBenchV6": 57.6,
            "GPQA-Diamond": 67.3,
            "IFEval": 86.0
        },
        {
            "key": 42,
            "model": "Qwen3-8B (Thinking)",
            "org": "Alibaba",
            "num": "8B",
            "time": "2025/4/29",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 56.5,
            "HLE": 5.5,
            "AIME2025": 66.2,
            "MMLU-Pro": 72.1,
            "LiveCodeBenchV6": 50.1,
            "GPQA-Diamond": 59.7,
            "IFEval": 85.6
        },
        {
            "key": 43,
            "model": "o3-2025-04-16",
            "org": "OpenAI",
            "num": "N/A",
            "time": "2025/4/16",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 72.8,
            "HLE": 21.3,
            "AIME2025": 85.1,
            "MMLU-Pro": 83.0,
            "LiveCodeBenchV6": 72.7,
            "GPQA-Diamond": 82.3,
            "IFEval": 92.2
        },
        {
            "key": 44,
            "model": "o4-mini-2025-04-16 (high)",
            "org": "OpenAI",
            "num": "N/A",
            "time": "2025/4/16",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 71.5,
            "HLE": 14.3,
            "AIME2025": 91.0,
            "MMLU-Pro": 82.4,
            "LiveCodeBenchV6": 71.0,
            "GPQA-Diamond": 77.4,
            "IFEval": 93.0
        },
        {
            "key": 45,
            "model": "GPT-4.1-20250414",
            "org": "OpenAI",
            "num": "N/A",
            "time": "2025/4/14",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 52.6,
            "HLE": 6.0,
            "AIME2025": 35.3,
            "MMLU-Pro": 81.0,
            "LiveCodeBenchV6": 37.2,
            "GPQA-Diamond": 68.1,
            "IFEval": 88.2
        },
        {
            "key": 46,
            "model": "Llama4-Maverick-17B-128E-Instruct",
            "org": "Meta",
            "num": "400B",
            "time": "2025/4/5",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 48.2,
            "HLE": 5.3,
            "AIME2025": 15.2,
            "MMLU-Pro": 79.6,
            "LiveCodeBenchV6": 33.2,
            "GPQA-Diamond": 69.1,
            "IFEval": 87.1
        },
        {
            "key": 47,
            "model": "Gemma-3-27B-it",
            "org": "Google",
            "num": "27B",
            "time": "2025/3/12",
            "update_time": "2026/3/9",
            "chat_or_base": "Chat",
            "Average": 42.1,
            "HLE": 4.2,
            "AIME2025": 22.4,
            "MMLU-Pro": 67.8,
            "LiveCodeBenchV6": 30.8,
            "GPQA-Diamond": 46.3,
            "IFEval": 81.0
        }
    ],
    "models": {
        "MiniCPM-SALA": {
            "desc": {
                "zh-CN": "MiniCPM SALA是第一个有效整合稀疏和线性注意力的大规模混合模型，用于百万Token上下文建模。",
                "en-US": "MiniCPM-SALA is the first large-scale hybrid model effectively integrating sparse and linear attention for million-token context modeling."
            },
            "org": "OpenBMB",
            "weight": [
                "openbmb/MiniCPM-SALA",
                "https://huggingface.co/openbmb/MiniCPM-SALA"
            ],
            "website-github": [
                "OpenBMB",
                "https://github.com/OpenBMB/MiniCPM/"
            ],
            "article": [
                "",
                ""
            ],
            "num": "9B",
            "time": "2026/2/11",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 1,
            "chat_or_base": "Chat"
        },
        "Qwen3.5-397B-A17B": {
            "desc": {
                "zh-CN": "Qwen3.5 是阿里近期推出的重要基础模型，在多模态融合、混合高效架构、规模化强化学习、多语言覆盖及训练基建上实现全面升级，性能与实用性大幅提升。",
                "en-US": "Qwen3.5 is a significant foundation model recently launched by Alibaba, featuring comprehensive upgrades in multimodal fusion, efficient hybrid architecture, scalable reinforcement learning, multilingual coverage and training infrastructure, with greatly improved performance and practicality."
            },
            "org": "Qwen",
            "weight": [
                "Qwen/Qwen3.5-397B-A17B",
                "https://huggingface.co/Qwen/Qwen3.5-397B-A17B"
            ],
            "website-github": [
                "Qwen",
                "https://github.com/QwenLM/Qwen3.5"
            ],
            "article": [
                "",
                ""
            ],
            "num": "397B",
            "time": "2026/2/16",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 2,
            "chat_or_base": "Chat"
        },
        "GLM-5": {
            "desc": {
                "zh-CN": "GLM-5 是面向复杂系统工程与长时序智能体任务推出的新一代模型，在保持长上下文能力的同时大幅降低部署成本，延续了通过模型缩放提升 AGI 智能效率的核心路径",
                "en-US": "GLM-5 is a new model launched for complex systems engineering and long-horizon agentic tasks, which significantly reduces deployment costs while maintaining long-context capabilities, and continues the core approach of improving AGI intelligence efficiency through model scaling."
            },
            "org": "Zhipu AI",
            "weight": [
                "zai-org/GLM-5",
                "https://huggingface.co/zai-org/GLM-5"
            ],
            "website-github": [
                "Zhipu AI",
                "https://z.ai/blog/glm-5"
            ],
            "article": [
                "",
                ""
            ],
            "num": "744B",
            "time": "2026/2/12",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 3,
            "chat_or_base": "Chat"
        },
        "MiniMax-M2.5": {
            "desc": {
                "zh-CN": "MiniMax-M2.5是在数十万复杂真实环境中经大规模强化学习训练，在编程、智能体工具使用等多项高价值任务上达到 SOTA 水平，推理与任务分解效率高、执行速度显著提升，且成本极低、性价比突出，有望催生全新智能体应用的前沿模型。",
                "en-US": "MiniMax-M2.5 is a cutting-edge model extensively trained via reinforcement learning in hundreds of thousands of complex real-world environments, achieving state-of-the-art performance in coding, agentic tool use, and many economically valuable tasks, with high reasoning efficiency, significantly improved execution speed, and extremely low cost, promising to enable innovative new agentic applications."
            },
            "org": "MiniMax",
            "weight": [
                "MiniMaxAI/MiniMax-M2.5",
                "https://huggingface.co/MiniMaxAI/MiniMax-M2.5"
            ],
            "website-github": [
                "MiniMax",
                "https://www.minimax.io/"
            ],
            "article": [
                "",
                ""
            ],
            "num": "230B",
            "time": "2026/2/12",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 4,
            "chat_or_base": "Chat"
        },
        "Kimi-K2.5": {
            "desc": {
                "zh-CN": "由月之暗面公司最新推出的模型，具备更强代码能力、更擅长通用Agent任务。所使用版本号为kimi-k2.5。",
                "en-US": "The latest model launched by Moonshot AI, featuring enhanced coding capabilities and superior performance in general Agent tasks, which adopts the version number kimi-k2.5."
            },
            "org": "Moonshot",
            "weight": [
                "moonshotai/Kimi-K2.5",
                ""
            ],
            "website-github": [
                "Moonshot",
                "https://www.moonshot.ai/"
            ],
            "article": [
                "",
                ""
            ],
            "num": "1000B",
            "time": "2026/1/27",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 5,
            "chat_or_base": "Chat"
        },
        "Step-3.5-Flash": {
            "desc": {
                "zh-CN": "Step-3.5-Flash adopts the MoE architecture, achieving a balance between inference ability and operational efficiency, significantly improving the response speed of the model in complex agent workflows.",
                "en-US": "Step-3.5-Flash采用了MoE架构，实现了推理能力与运行效率的平衡，显著提升了模型在复杂Agent工作流中的响应速度。"
            },
            "org": "StepFun",
            "weight": [
                "stepfun-ai/Step-3.5-Flash",
                "https://huggingface.co/stepfun-ai/Step-3.5-Flash"
            ],
            "website-github": [
                "StepFun",
                "https://static.stepfun.com/blog/step-3.5-flash/"
            ],
            "article": [
                "",
                ""
            ],
            "num": "196B",
            "time": "2026/2/2",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 6,
            "chat_or_base": "Chat"
        },
        "TeleChat-thinking": {
            "desc": {
                "zh-CN": "中国电信人工智能研究院（TeleAI）推出的大语言模型。",
                "en-US": "The large language model launched by TeleAI."
            },
            "org": "TeleAI",
            "weight": [
                "",
                ""
            ],
            "website-github": [
                "TeleAI",
                "https://www.teleai.com.cn/"
            ],
            "article": [
                "",
                ""
            ],
            "num": "N/A",
            "time": "2025/12/29",
            "update_time": "2026/3/9",
            "release": "API",
            "rank": 7,
            "chat_or_base": "Chat"
        },
        "GLM-4.7": {
            "desc": {
                "zh-CN": "GLM发布的新一代GLM-4.7系列开源推理模型。",
                "en-US": "The new GLM-4.7 reasoning model series released by GLM."
            },
            "org": "Zhipu AI",
            "weight": [
                "zai-org/GLM-4.7",
                "https://huggingface.co/zai-org/GLM-4.7"
            ],
            "website-github": [
                "Zhipu AI",
                "https://z.ai/model-api"
            ],
            "article": [
                "",
                ""
            ],
            "num": "355B",
            "time": "2025/12/23",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 8,
            "chat_or_base": "Chat"
        },
        "MiMo-V2-Flash": {
            "desc": {
                "zh-CN": "MiMo-V2-Flash is an MoE language model, designed for high-speed reasoning and agentic workflows, it utilizes a novel hybrid attention architecture and MTP to achieve state-of-the-art performance while significantly reducing inference costs.",
                "en-US": "MiMo-V2-Flash是一个MoE语言模型，专为高速推理和代理工作流程而设计，利用新颖的混合注意力架构和MTP来实现最先进的性能，同时显著降低推理成本。"
            },
            "org": "Xiaomi",
            "weight": [
                "XiaomiMiMo/MiMo-V2-Flash",
                "https://huggingface.co/XiaomiMiMo/MiMo-V2-Flash"
            ],
            "website-github": [
                "Xiaomi",
                "https://platform.xiaomimimo.com/#/docs/welcome"
            ],
            "article": [
                "",
                ""
            ],
            "num": "309B",
            "time": "2025/12/17",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 9,
            "chat_or_base": "Chat"
        },
        "DeepSeek-V3.2": {
            "desc": {
                "zh-CN": "DeepSeek最新推出的V3.2模型正式版，该模型在保持卓越推理能力和智能体性能的同时，实现了高计算效率的平衡。",
                "en-US": "The official release of DeepSeek’s newly launched V3.2 model has just been announced. While maintaining outstanding reasoning capabilities and agent performance, the model also strikes a balance with high computational efficiency."
            },
            "org": "DeepSeek",
            "weight": [
                "deepseek-ai/DeepSeek-V3.2",
                "https://huggingface.co/deepseek-ai/DeepSeek-V3.2"
            ],
            "website-github": [
                "DeepSeek",
                "https://huggingface.co/deepseek-ai/DeepSeek-V3.2"
            ],
            "article": [
                "",
                ""
            ],
            "num": "671B",
            "time": "2025/12/1",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 10,
            "chat_or_base": "Chat"
        },
        "DeepSeek-V3.2-Speciale": {
            "desc": {
                "zh-CN": "DeepSeekV3.2的高计算版本，具有更强的计算和推理能力。",
                "en-US": "The high-compute version of DeepSeekV3.2, offering stronger computational and reasoning capabilities."
            },
            "org": "DeepSeek",
            "weight": [
                "deepseek-ai/DeepSeek-V3.2-Speciale",
                "https://huggingface.co/deepseek-ai/DeepSeek-V3.2-Speciale"
            ],
            "website-github": [
                "DeepSeek",
                "https://huggingface.co/deepseek-ai/DeepSeek-V3.2-Speciale"
            ],
            "article": [
                "",
                ""
            ],
            "num": "671B",
            "time": "2025/12/1",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 11,
            "chat_or_base": "Chat"
        },
        "Gemini-3-Pro-Preview": {
            "desc": {
                "zh-CN": "gemini3-pro模型的预览版本。",
                "en-US": "The preview version of the Gemini 3 pro."
            },
            "org": "Google",
            "weight": [
                "",
                ""
            ],
            "website-github": [
                "Google",
                "https://developers.googleblog.com/new-gemini-api-updates-for-gemini-3/"
            ],
            "article": [
                "",
                ""
            ],
            "num": "N/A",
            "time": "2025/11/19",
            "update_time": "2026/3/9",
            "release": "API",
            "rank": 12,
            "chat_or_base": "Chat"
        },
        "Kimi-K2-Thinking": {
            "desc": {
                "zh-CN": "由月之暗面公司最新推出的深度思考模型，具备更强代码能力、更擅长通用Agent任务的MoE架构模型。",
                "en-US": "The latest deep reasoning model launched by Moonshot AI, featuring enhanced coding capabilities and superior performance in general Agent tasks, built on an MoE architecture."
            },
            "org": "Moonshot",
            "weight": [
                "moonshotai/Kimi-K2-Thinking",
                "https://huggingface.co/moonshotai/Kimi-K2-Thinking"
            ],
            "website-github": [
                "Moonshot",
                "https://moonshotai.github.io/Kimi-K2/thinking.html"
            ],
            "article": [
                "",
                ""
            ],
            "num": "1000B",
            "time": "2025/11/6",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 13,
            "chat_or_base": "Chat"
        },
        "MiniMax-M2": {
            "desc": {
                "zh-CN": "Minimax最新发布的旗舰模型，具有卓越的通用能力、先进的编程能力、强大的智能体性能以及高效的架构设计。",
                "en-US": "The latest flagship model released by Minimax features excellent general capabilities, advanced programming capabilities, powerful agent performance, and efficient architecture design."
            },
            "org": "MiniMax",
            "weight": [
                "MiniMaxAI/MiniMax-M2",
                "https://huggingface.co/MiniMaxAI/MiniMax-M2"
            ],
            "website-github": [
                "MiniMax",
                "https://www.minimaxi.com/platform"
            ],
            "article": [
                "",
                ""
            ],
            "num": "230B",
            "time": "2025/10/25",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 14,
            "chat_or_base": "Chat"
        },
        "JT-Think": {
            "desc": {
                "zh-CN": "中国移动推出的闭源深度思考模型。",
                "en-US": "A close-source deep reasoning model launched by China Mobile."
            },
            "org": "China Mobile",
            "weight": [
                "",
                ""
            ],
            "website-github": [
                "China Mobile",
                "https://jiutian.10086.cn/portal/#/home"
            ],
            "article": [
                "",
                ""
            ],
            "num": "N/A",
            "time": "2025/10/22",
            "update_time": "2026/3/9",
            "release": "API",
            "rank": 15,
            "chat_or_base": "Chat"
        },
        "GLM-4.6": {
            "desc": {
                "zh-CN": "GLM发布的新一代GLM-4.6系列开源推理模型。",
                "en-US": "The new GLM-4.6 reasoning model series released by GLM."
            },
            "org": "Zhipu AI",
            "weight": [
                "zai-org/GLM-4.6",
                "https://huggingface.co/zai-org/GLM-4.6"
            ],
            "website-github": [
                "Zhipu AI",
                "https://z.ai/blog/glm-4.6"
            ],
            "article": [
                "",
                ""
            ],
            "num": "355B",
            "time": "2025/9/30",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 16,
            "chat_or_base": "Chat"
        },
        "Qwen3-Next-80B-A3B-Thinking": {
            "desc": {
                "zh-CN": "Qwen模型中最新发布的Qwen3-Next系列，通过创新的模型架构改进Scaling效率。",
                "en-US": "The latest released Qwen3-Next series in Qwen models, improving scaling efficiency through innovative model architecture."
            },
            "org": "Alibaba",
            "weight": [
                "Qwen/Qwen3-Next-80B-A3B-Thinking",
                "https://huggingface.co/Qwen/Qwen3-Next-80B-A3B-Thinking"
            ],
            "website-github": [
                "Qwen3",
                "https://github.com/QwenLM/Qwen3"
            ],
            "article": [
                "",
                ""
            ],
            "num": "80B",
            "time": "2025/9/12",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 17,
            "chat_or_base": "Chat"
        },
        "Qwen3-Next-80B-A3B-Instruct": {
            "desc": {
                "zh-CN": "Qwen模型中最新发布的Qwen3-Next系列，通过创新的模型架构改进Scaling效率。",
                "en-US": "The latest released Qwen3-Next series in Qwen models, improving scaling efficiency through innovative model architecture."
            },
            "org": "Alibaba",
            "weight": [
                "Qwen/Qwen3-Next-80B-A3B-Instruct",
                "https://huggingface.co/Qwen/Qwen3-Next-80B-A3B-Instruct"
            ],
            "website-github": [
                "Qwen3",
                "https://github.com/QwenLM/Qwen3"
            ],
            "article": [
                "",
                ""
            ],
            "num": "80B",
            "time": "2025/9/12",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 18,
            "chat_or_base": "Chat"
        },
        "ERNIE-4.5-21B-A3B-Thinking": {
            "desc": {
                "zh-CN": "百度在文心一言4.5系列的基础上发布的新模型，其思维能力得到扩展，推理质量和深度得到提升。",
                "en-US": "The new model released by Baidu on the basis of ERNIE-4.5 series, which has expanded its thinking ability and improved the quality and depth of reasoning."
            },
            "org": "Baidu",
            "weight": [
                "baidu/ERNIE-4.5-21B-A3B-Thinking",
                "https://huggingface.co/baidu/ERNIE-4.5-21B-A3B-Thinking"
            ],
            "website-github": [
                "Baidu",
                "https://yiyan.baidu.com/blog/posts/ernie4.5"
            ],
            "article": [
                "",
                ""
            ],
            "num": "21B",
            "time": "2025/9/9",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 19,
            "chat_or_base": "Chat"
        },
        "LongCat-Flash-Chat": {
            "desc": {
                "zh-CN": "美团推出的一款名为 LongCat-Flash 的非推理开源大模型，以极高的推理效率著称。",
                "en-US": "An open-sourced non-reasoning LLM named LongCat-Flash launched by Meituan, known for its extremely high inference efficiency."
            },
            "org": "Meituan",
            "weight": [
                "meituan-longcat/LongCat-Flash-Chat",
                "https://huggingface.co/meituan-longcat/LongCat-Flash-Chat"
            ],
            "website-github": [
                "LongCat",
                "https://longcat.chat/"
            ],
            "article": [
                "",
                ""
            ],
            "num": "562B",
            "time": "2025/9/1",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 20,
            "chat_or_base": "Chat"
        },
        "Seed-OSS-36B-Instruct": {
            "desc": {
                "zh-CN": "Seed-OSS是字节跳动开发的一系列开源大型语言模型，旨在提供强大的长上下文、推理、代理和通用功能，以及通用的开发人员友好功能。尽管仅使用12T令牌进行训练，但Seed-OSS在几个流行的开放基准测试中取得了出色的性能。",
                "en-US": "Seed-OSS is a series of open-source large language models developed by ByteDance's Seed Team, designed for powerful long-context, reasoning, agent and general capabilities, and versatile developer-friendly features. Although trained with only 12T tokens, Seed-OSS achieves excellent performance on several popular open benchmarks."
            },
            "org": "ByteDance",
            "weight": [
                "ByteDance-Seed/Seed-OSS-36B-Instruct",
                "https://huggingface.co/ByteDance-Seed/Seed-OSS-36B-Instruct"
            ],
            "website-github": [
                "ByteDance-Seed",
                "https://github.com/ByteDance-Seed/seed-oss"
            ],
            "article": [
                "",
                ""
            ],
            "num": "36B",
            "time": "2025/8/20",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 21,
            "chat_or_base": "Chat"
        },
        "Qwen3-4B-Thinking-2507": {
            "desc": {
                "zh-CN": "Qwen3是Qwen系列中最新一代的大型语言模型，提供了一套全面的密集模型和MoE模型。",
                "en-US": "Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. "
            },
            "org": "Alibaba",
            "weight": [
                "Qwen/Qwen3-4B-Thinking-2507",
                "https://huggingface.co/Qwen/Qwen3-4B-Thinking-2507"
            ],
            "website-github": [
                "Qwen3",
                "https://github.com/QwenLM/Qwen3"
            ],
            "article": [
                "",
                ""
            ],
            "num": "4B",
            "time": "2025/8/7",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 22,
            "chat_or_base": "Chat"
        },
        "Qwen3-4B-Instruct-2507": {
            "desc": {
                "zh-CN": "Qwen3是Qwen系列中最新一代的大型语言模型，提供了一套全面的密集模型和MoE模型。",
                "en-US": "Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. "
            },
            "org": "Alibaba",
            "weight": [
                "Qwen/Qwen3-4B-Instruct-2507",
                "https://huggingface.co/Qwen/Qwen3-4B-Instruct-2507"
            ],
            "website-github": [
                "Qwen3",
                "https://github.com/QwenLM/Qwen3"
            ],
            "article": [
                "",
                ""
            ],
            "num": "4B",
            "time": "2025/8/7",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 23,
            "chat_or_base": "Chat"
        },
        "GPT-5-2025-08-07 (high)": {
            "desc": {
                "zh-CN": "OpenAI官方最新推出的备受外界期待的、性能更强的新一代人工智能模型GPT-5。",
                "en-US": "The latest, highly anticipated and powerful next-generation LLM named GPT-5 that officially launched by OpenAI."
            },
            "org": "OpenAI",
            "weight": [
                "",
                ""
            ],
            "website-github": [
                "OpenAI",
                "https://openai.com/index/introducing-gpt-5/"
            ],
            "article": [
                "",
                ""
            ],
            "num": "N/A",
            "time": "2025/8/7",
            "update_time": "2026/3/9",
            "release": "API",
            "rank": 24,
            "chat_or_base": "Chat"
        },
        "GPT-5-mini-2025-08-07 (high)": {
            "desc": {
                "zh-CN": "OpenAI官方最新推出的备受外界期待的、性能更强的新一代人工智能模型GPT-5。",
                "en-US": "The latest, highly anticipated and powerful next-generation LLM named GPT-5 that officially launched by OpenAI."
            },
            "org": "OpenAI",
            "weight": [
                "",
                ""
            ],
            "website-github": [
                "OpenAI",
                "https://openai.com/index/introducing-gpt-5/"
            ],
            "article": [
                "",
                ""
            ],
            "num": "N/A",
            "time": "2025/8/7",
            "update_time": "2026/3/9",
            "release": "API",
            "rank": 25,
            "chat_or_base": "Chat"
        },
        "GPT-5-nano-2025-08-07 (high)": {
            "desc": {
                "zh-CN": "OpenAI官方最新推出的备受外界期待的、性能更强的新一代人工智能模型GPT-5。",
                "en-US": "The latest, highly anticipated and powerful next-generation LLM named GPT-5 that officially launched by OpenAI."
            },
            "org": "OpenAI",
            "weight": [
                "",
                ""
            ],
            "website-github": [
                "OpenAI",
                "https://openai.com/index/introducing-gpt-5/"
            ],
            "article": [
                "",
                ""
            ],
            "num": "N/A",
            "time": "2025/8/7",
            "update_time": "2026/3/9",
            "release": "API",
            "rank": 26,
            "chat_or_base": "Chat"
        },
        "gpt-oss-120b (high)": {
            "desc": {
                "zh-CN": "由OpenAI发布的近年来首个开源模型系列gpt-oss。",
                "en-US": "The GPT-OSS series, the first open-source model family released by OpenAI in recent years."
            },
            "org": "OpenAI",
            "weight": [
                "openai/gpt-oss-120b",
                "https://huggingface.co/openai/gpt-oss-120b"
            ],
            "website-github": [
                "OpenAI",
                "https://openai.com/index/introducing-gpt-oss/"
            ],
            "article": [
                "",
                ""
            ],
            "num": "120B",
            "time": "2025/8/5",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 27,
            "chat_or_base": "Chat"
        },
        "gpt-oss-20b (high)": {
            "desc": {
                "zh-CN": "由OpenAI发布的近年来首个开源模型系列gpt-oss。",
                "en-US": "The GPT-OSS series, the first open-source model family released by OpenAI in recent years."
            },
            "org": "OpenAI",
            "weight": [
                "openai/gpt-oss-20b",
                "https://huggingface.co/openai/gpt-oss-20b"
            ],
            "website-github": [
                "OpenAI",
                "https://openai.com/index/introducing-gpt-oss/"
            ],
            "article": [
                "",
                ""
            ],
            "num": "20B",
            "time": "2025/8/5",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 28,
            "chat_or_base": "Chat"
        },
        "Qwen3-30B-A3B-Thinking-2507": {
            "desc": {
                "zh-CN": "Qwen3是Qwen系列中最新一代的大型语言模型，提供了一套全面的密集模型和MoE模型。",
                "en-US": "Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. "
            },
            "org": "Alibaba",
            "weight": [
                "Qwen/Qwen3-30B-A3B-Thinking-2507",
                "https://huggingface.co/Qwen/Qwen3-30B-A3B-Thinking-2507"
            ],
            "website-github": [
                "Qwen3",
                "https://github.com/QwenLM/Qwen3"
            ],
            "article": [
                "",
                ""
            ],
            "num": "30B",
            "time": "2025/7/30",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 29,
            "chat_or_base": "Chat"
        },
        "Qwen3-30B-A3B-Instruct-2507": {
            "desc": {
                "zh-CN": "Qwen3是Qwen系列中最新一代的大型语言模型，提供了一套全面的密集模型和MoE模型。",
                "en-US": "Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. "
            },
            "org": "Alibaba",
            "weight": [
                "Qwen/Qwen3-30B-A3B-Instruct-2507",
                "https://huggingface.co/Qwen/Qwen3-30B-A3B-Instruct-2507"
            ],
            "website-github": [
                "Qwen3",
                "https://github.com/QwenLM/Qwen3"
            ],
            "article": [
                "",
                ""
            ],
            "num": "30B",
            "time": "2025/7/30",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 30,
            "chat_or_base": "Chat"
        },
        "GLM-4.5": {
            "desc": {
                "zh-CN": "GLM发布的新一代GLM-4.5系列开源推理模型。",
                "en-US": "The new GLM-4.5 reasoning model series released by GLM."
            },
            "org": "Zhipu AI",
            "weight": [
                "zai-org/GLM-4.5",
                "https://huggingface.co/zai-org/GLM-4.5"
            ],
            "website-github": [
                "GLM-4.5",
                "https://z.ai/blog/glm-4.5"
            ],
            "article": [
                "",
                ""
            ],
            "num": "358B",
            "time": "2025/7/29",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 31,
            "chat_or_base": "Chat"
        },
        "Qwen3-235B-A22B-Thinking-2507": {
            "desc": {
                "zh-CN": "Qwen3是Qwen系列中最新一代的大型语言模型，提供了一套全面的密集模型和MoE模型。",
                "en-US": "Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. "
            },
            "org": "Alibaba",
            "weight": [
                "Qwen/Qwen3-235B-A22B-Thinking-2507",
                "https://huggingface.co/Qwen/Qwen3-235B-A22B-Thinking-2507"
            ],
            "website-github": [
                "Qwen3",
                "https://github.com/QwenLM/Qwen3"
            ],
            "article": [
                "report",
                "https://qwenlm.github.io/zh/blog/qwen3/"
            ],
            "num": "235B",
            "time": "2025/7/25",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 32,
            "chat_or_base": "Chat"
        },
        "Qwen3-235B-A22B-Instruct-2507": {
            "desc": {
                "zh-CN": "Qwen3是Qwen系列中最新一代的大型语言模型，提供了一套全面的密集模型和MoE模型。",
                "en-US": "Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. "
            },
            "org": "Alibaba",
            "weight": [
                "Qwen/Qwen3-235B-A22B-Instruct-2507",
                "https://huggingface.co/Qwen/Qwen3-235B-A22B-Instruct-2507"
            ],
            "website-github": [
                "Qwen3",
                "https://github.com/QwenLM/Qwen3"
            ],
            "article": [
                "report",
                "https://qwenlm.github.io/zh/blog/qwen3/"
            ],
            "num": "235B",
            "time": "2025/7/22",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 33,
            "chat_or_base": "Chat"
        },
        "Kimi-K2-Instruct": {
            "desc": {
                "zh-CN": "由月之暗面公司最新推出的模型，具备更强代码能力、更擅长通用Agent任务的MoE架构模型。",
                "en-US": "The latest model launched by Moonshot AI, featuring enhanced coding capabilities and superior performance in general Agent tasks, built on an MoE architecture."
            },
            "org": "Moonshot",
            "weight": [
                "moonshotai/Kimi-K2-Instruct",
                "https://huggingface.co/moonshotai/Kimi-K2-Instruct"
            ],
            "website-github": [
                "Moonshot",
                "https://moonshotai.github.io/Kimi-K2/"
            ],
            "article": [
                "",
                ""
            ],
            "num": "1000B",
            "time": "2025/7/11",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 34,
            "chat_or_base": "Chat"
        },
        "ERNIE-4.5-300B-A47B-PT": {
            "desc": {
                "zh-CN": "百度文心一言系列模型的最新开源版本。",
                "en-US": "The latest open source version of Baidu ERNIE series LLMs."
            },
            "org": "Baidu",
            "weight": [
                "baidu/ERNIE-4.5-300B-A47B-PT",
                "https://huggingface.co/baidu/ERNIE-4.5-300B-A47B-PT"
            ],
            "website-github": [
                "Baidu",
                "https://yiyan.baidu.com/blog/posts/ernie4.5"
            ],
            "article": [
                "",
                ""
            ],
            "num": "300B",
            "time": "2025/6/30",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 35,
            "chat_or_base": "Chat"
        },
        "Hunyuan-A13B-Instruct": {
            "desc": {
                "zh-CN": "腾讯最新发布的MOE模型。",
                "en-US": "The latest MOE model released by Tencent."
            },
            "org": "Tencent",
            "weight": [
                "tencent/Hunyuan-A13B-Instruct",
                "https://huggingface.co/tencent/Hunyuan-A13B-Instruct"
            ],
            "website-github": [
                "Tencent",
                "https://hunyuan.tencent.com/"
            ],
            "article": [
                "",
                ""
            ],
            "num": "80B",
            "time": "2025/6/27",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 36,
            "chat_or_base": "Chat"
        },
        "Finix-P1-32B (Thinking)": {
            "desc": {
                "zh-CN": "由蚂蚁集团发布的金融领域大模型。",
                "en-US": "The financial domain LLM released by Ant Group"
            },
            "org": "Ant Group",
            "weight": [
                "",
                ""
            ],
            "website-github": [
                "Ant Group",
                "https://antfinix.alipay.com/"
            ],
            "article": [
                "",
                ""
            ],
            "num": "N/A",
            "time": "2025/6/24",
            "update_time": "2026/3/9",
            "release": "API",
            "rank": 37,
            "chat_or_base": "Chat"
        },
        "Gemini-2.5-Pro": {
            "desc": {
                "zh-CN": "gemini2.5-pro模型的正式发布版本。",
                "en-US": "The official release version of the Gemini 2.5 pro."
            },
            "org": "Google",
            "weight": [
                "",
                ""
            ],
            "website-github": [
                "Google",
                "https://developers.googleblog.com/en/gemini-2-5-thinking-model-updates/"
            ],
            "article": [
                "",
                ""
            ],
            "num": "N/A",
            "time": "2025/6/18",
            "update_time": "2026/3/9",
            "release": "API",
            "rank": 38,
            "chat_or_base": "Chat"
        },
        "doubao-seed-1.6-thinking-250615": {
            "desc": {
                "zh-CN": "字节豆包推出的seed系列模型的最新版，支持思考模式。",
                "en-US": "The latest version of the seed series model launched by ByteDance, whitch supports the thinking mode."
            },
            "org": "ByteDance",
            "weight": [
                "",
                ""
            ],
            "website-github": [
                "Doubao",
                "https://console.volcengine.com/ark"
            ],
            "article": [
                "",
                ""
            ],
            "num": "N/A",
            "time": "2025/6/11",
            "update_time": "2026/3/9",
            "release": "API",
            "rank": 39,
            "chat_or_base": "Chat"
        },
        "DeepSeek-R1-0528": {
            "desc": {
                "zh-CN": "Deepseek R1的最新版本。",
                "en-US": "The latest version of Deepseek R1."
            },
            "org": "DeepSeek",
            "weight": [
                "deepseek-ai/DeepSeek-R1-0528",
                "https://huggingface.co/deepseek-ai/DeepSeek-R1-0528"
            ],
            "website-github": [
                "DeepSeek",
                "https://github.com/deepseek-ai/DeepSeek-R1"
            ],
            "article": [
                "",
                ""
            ],
            "num": "671B",
            "time": "2025/5/28",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 40,
            "chat_or_base": "Chat"
        },
        "Claude Sonnet 4 (Thinking)": {
            "desc": {
                "zh-CN": "Anthropic最新发布的混合推理模型系列，支持更好的代码编写、工件调用等功能。",
                "en-US": "The latest hybrid inference model series released by Anthropic, which support better coding, function calling, and other features."
            },
            "org": "Anthropic",
            "weight": [
                "",
                ""
            ],
            "website-github": [
                "Anthropic",
                "https://www.anthropic.com/claude/sonnet"
            ],
            "article": [
                "",
                ""
            ],
            "num": "N/A",
            "time": "2025/5/22",
            "update_time": "2026/3/9",
            "release": "API",
            "rank": 41,
            "chat_or_base": "Chat"
        },
        "Qwen3-32B (Thinking)": {
            "desc": {
                "zh-CN": "Qwen3是Qwen系列中最新一代的大型语言模型，提供了一套全面的密集模型和MoE模型。",
                "en-US": "Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. "
            },
            "org": "Alibaba",
            "weight": [
                "Qwen/Qwen3-32B",
                "https://huggingface.co/Qwen/Qwen3-32B"
            ],
            "website-github": [
                "Qwen3",
                "https://github.com/QwenLM/Qwen3"
            ],
            "article": [
                "report",
                "https://qwenlm.github.io/zh/blog/qwen3/"
            ],
            "num": "32B",
            "time": "2025/4/29",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 42,
            "chat_or_base": "Chat"
        },
        "Qwen3-8B (Thinking)": {
            "desc": {
                "zh-CN": "Qwen3是Qwen系列中最新一代的大型语言模型，提供了一套全面的密集模型和MoE模型。",
                "en-US": "Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. "
            },
            "org": "Alibaba",
            "weight": [
                "Qwen/Qwen3-8B",
                "https://huggingface.co/Qwen/Qwen3-8B"
            ],
            "website-github": [
                "Qwen3",
                "https://github.com/QwenLM/Qwen3"
            ],
            "article": [
                "report",
                "https://qwenlm.github.io/zh/blog/qwen3/"
            ],
            "num": "8B",
            "time": "2025/4/29",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 43,
            "chat_or_base": "Chat"
        },
        "o3-2025-04-16": {
            "desc": {
                "zh-CN": "o3是openai发布的最新深度思考模型，其基线指标大幅超越上一代模型。",
                "en-US": "O3 is the latest reasoning model released by OpenAI, with baseline metrics significantly surpassing previous models."
            },
            "org": "OpenAI",
            "weight": [
                "",
                ""
            ],
            "website-github": [
                "OpenAI",
                "https://openai.com/index/introducing-o3-and-o4-mini/"
            ],
            "article": [
                "",
                ""
            ],
            "num": "N/A",
            "time": "2025/4/16",
            "update_time": "2026/3/9",
            "release": "API",
            "rank": 44,
            "chat_or_base": "Chat"
        },
        "o4-mini-2025-04-16 (high)": {
            "desc": {
                "zh-CN": "high模式下的o4-mini，响应时间更长，具有更强的推理能力。",
                "en-US": "o4-mini in high mode, which has a longer response time and stronger reasoning ability."
            },
            "org": "OpenAI",
            "weight": [
                "",
                ""
            ],
            "website-github": [
                "OpenAI",
                "https://openai.com/index/introducing-o3-and-o4-mini/"
            ],
            "article": [
                "",
                ""
            ],
            "num": "N/A",
            "time": "2025/4/16",
            "update_time": "2026/3/9",
            "release": "API",
            "rank": 45,
            "chat_or_base": "Chat"
        },
        "GPT-4.1-20250414": {
            "desc": {
                "zh-CN": "由OpenAI推出的GPT-4.1系列模型，在编码、指令遵循和长上下文方面有重大改进。",
                "en-US": "The GPT-4.1 series models launched by OpenAI, which have significant improvements in encoding, instruction adherence, and long context."
            },
            "org": "OpenAI",
            "weight": [
                "",
                ""
            ],
            "website-github": [
                "OpenAI",
                "https://platform.openai.com/docs/models/gpt-4.1"
            ],
            "article": [
                "",
                ""
            ],
            "num": "N/A",
            "time": "2025/4/14",
            "update_time": "2026/3/9",
            "release": "API",
            "rank": 46,
            "chat_or_base": "Chat"
        },
        "Llama4-Maverick-17B-128E-Instruct": {
            "desc": {
                "zh-CN": "Llama 4模型是自回归语言模型，它使用混合专家（MoE）架构，并结合了原生多模态的早期融合",
                "en-US": "The Llama 4 models are auto-regressive language models that use a mixture-of-experts (MoE) architecture and incorporate early fusion for native multimodality"
            },
            "org": "Meta",
            "weight": [
                "meta-llama/Llama-4-Maverick-17B-128E-Instruct",
                "https://huggingface.co/meta-llama/Llama-4-Maverick-17B-128E-Instruct"
            ],
            "website-github": [
                "Llama4-Maverick",
                "https://www.llama.com/"
            ],
            "article": [
                "",
                ""
            ],
            "num": "400B",
            "time": "2025/4/5",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 47,
            "chat_or_base": "Chat"
        },
        "Gemma-3-27B-it": {
            "desc": {
                "zh-CN": "Gemma 是来自谷歌的一系列轻量级、最先进的开源模型家族，这些模型基于创建Gemini模型相同的研究和技术构建而成。",
                "en-US": "Gemma is a family of lightweight, state-of-the-art open models from Google, built from the same research and technology used to create the Gemini models."
            },
            "org": "Google",
            "weight": [
                "google/gemma-3-27b-it",
                "https://huggingface.co/google/gemma-3-27b-it"
            ],
            "website-github": [
                "Google",
                "https://ai.google.dev/gemma/docs"
            ],
            "article": [
                "",
                ""
            ],
            "num": "27B",
            "time": "2025/3/12",
            "update_time": "2026/3/9",
            "release": "OpenSource",
            "rank": 48,
            "chat_or_base": "Chat"
        }
    },
    "globalData": {
        "说明": "如果中英文环境不同，用models.ChatGPT.desc的写法，如果中英文相同，可以直接用字符串",
        "description": {
            "zh-CN": "OpenCompass 学术开源评测集目前聚焦大语言模型综合推理能力。\n- 目前选取的数据集包括 综合知识推理 (MMLU-Pro /GPQA-Diamond(repeat-4) /HLE) ,数学推理 (AIME2025(repeat-32)) ,代码生成 (LiveCodeBench(repeat-6)) ,指令跟随 (IFEval)。\n- 评测集会以月度为单位进行持续更新，纳入社区最新的研究进展。\n- 当前模型评测以对话模型为主，不定期更新社区最新模型，相关提示词和复现脚本参考OpenCompass代码仓库。\n- 模型评测结果会随着工具链的迭代同步更新。\n- OpenCompass项目中提供了复现学术榜单的配置文件，具体请查阅官方文档。",
            "en-US": "The CompassAcademic currently focuses on the comprehensive reasoning abilities of LLMs.\n- The datasets selected so far include General Knowledge Reasoning (MMLU-Pro /GPQA-Diamond(repeat-4) /HLE), Mathematical Reasoning (AIME2025(repeat-32)), Code Completion (LiveCodeBench(repeat-6)), and Instruction Following (IFEval).\n- The benchmark is updated monthly to incorporate the latest research advancements from the community.\n- Currently, the evaluation primarily targets chat models, with updates featuring the latest community models at irregular intervals. Prompts and reproduction scripts can be found in the OpenCompass toolkit.\n- The model evaluation results will be updated synchronously with the iteration of the toolchain. \n- OpenCompass provides configurations for reproducing this leaderboard. Please refer to the documentation for details."
        },
        "OverallTable": [
            {
                "key": 0,
                "model": "MiniCPM-SALA",
                "org": "OpenBMB",
                "num": "9B",
                "time": "2026/2/11",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 58.7,
                "HLE": 9.3,
                "AIME2025": 81.0,
                "MMLU-Pro": 72.0,
                "LiveCodeBenchV6": 54.7,
                "GPQA-Diamond": 58.2,
                "IFEval": 76.9
            },
            {
                "key": 1,
                "model": "Qwen3.5-397B-A17B",
                "org": "Qwen",
                "num": "397B",
                "time": "2026/2/16",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 78.4,
                "HLE": 27.5,
                "AIME2025": 92.3,
                "MMLU-Pro": 87.6,
                "LiveCodeBenchV6": 83.0,
                "GPQA-Diamond": 88.4,
                "IFEval": 91.5
            },
            {
                "key": 2,
                "model": "GLM-5",
                "org": "Zhipu AI",
                "num": "744B",
                "time": "2026/2/12",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 79.0,
                "HLE": 28.1,
                "AIME2025": 95.8,
                "MMLU-Pro": 85.2,
                "LiveCodeBenchV6": 86.2,
                "GPQA-Diamond": 85.3,
                "IFEval": 93.2
            },
            {
                "key": 3,
                "model": "MiniMax-M2.5",
                "org": "MiniMax",
                "num": "230B",
                "time": "2026/2/12",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 73.3,
                "HLE": 22.2,
                "AIME2025": 86.2,
                "MMLU-Pro": 81.7,
                "LiveCodeBenchV6": 73.6,
                "GPQA-Diamond": 84.6,
                "IFEval": 91.1
            },
            {
                "key": 4,
                "model": "Kimi-K2.5",
                "org": "Moonshot",
                "num": "1000B",
                "time": "2026/1/27",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 78.2,
                "HLE": 28.6,
                "AIME2025": 91.9,
                "MMLU-Pro": 86.2,
                "LiveCodeBenchV6": 80.6,
                "GPQA-Diamond": 88.1,
                "IFEval": 93.9
            },
            {
                "key": 5,
                "model": "Step-3.5-Flash",
                "org": "StepFun",
                "num": "196B",
                "time": "2026/2/2",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 76.9,
                "HLE": 21.6,
                "AIME2025": 95.7,
                "MMLU-Pro": 83.5,
                "LiveCodeBenchV6": 83.9,
                "GPQA-Diamond": 83.7,
                "IFEval": 93.2
            },
            {
                "key": 6,
                "model": "TeleChat-thinking",
                "org": "TeleAI",
                "num": "N/A",
                "time": "2025/12/29",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 71.8,
                "HLE": 20.3,
                "AIME2025": 79.0,
                "MMLU-Pro": 84.1,
                "LiveCodeBenchV6": 71.7,
                "GPQA-Diamond": 81.4,
                "IFEval": 94.3
            },
            {
                "key": 7,
                "model": "GLM-4.7",
                "org": "Zhipu AI",
                "num": "355B",
                "time": "2025/12/23",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 77.6,
                "HLE": 25.4,
                "AIME2025": 95.4,
                "MMLU-Pro": 84.0,
                "LiveCodeBenchV6": 83.8,
                "GPQA-Diamond": 86.9,
                "IFEval": 90.2
            },
            {
                "key": 8,
                "model": "MiMo-V2-Flash",
                "org": "Xiaomi",
                "num": "309B",
                "time": "2025/12/17",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 73.3,
                "HLE": 20.5,
                "AIME2025": 92.9,
                "MMLU-Pro": 83.1,
                "LiveCodeBenchV6": 71.8,
                "GPQA-Diamond": 82.1,
                "IFEval": 89.5
            },
            {
                "key": 9,
                "model": "DeepSeek-V3.2",
                "org": "DeepSeek",
                "num": "671B",
                "time": "2025/12/1",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 75.3,
                "HLE": 23.2,
                "AIME2025": 93.0,
                "MMLU-Pro": 85.8,
                "LiveCodeBenchV6": 75.4,
                "GPQA-Diamond": 84.6,
                "IFEval": 89.7
            },
            {
                "key": 10,
                "model": "DeepSeek-V3.2-Speciale",
                "org": "DeepSeek",
                "num": "671B",
                "time": "2025/12/1",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 78.2,
                "HLE": 28.6,
                "AIME2025": 96.0,
                "MMLU-Pro": 85.5,
                "LiveCodeBenchV6": 80.9,
                "GPQA-Diamond": 86.7,
                "IFEval": 91.7
            },
            {
                "key": 11,
                "model": "Gemini-3-Pro-Preview",
                "org": "Google",
                "num": "N/A",
                "time": "2025/11/19",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 81.3,
                "HLE": 38.0,
                "AIME2025": 93.4,
                "MMLU-Pro": 89.3,
                "LiveCodeBenchV6": 82.9,
                "GPQA-Diamond": 91.5,
                "IFEval": 92.8
            },
            {
                "key": 12,
                "model": "Kimi-K2-Thinking",
                "org": "Moonshot",
                "num": "1000B",
                "time": "2025/11/6",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 75.3,
                "HLE": 21.3,
                "AIME2025": 94.1,
                "MMLU-Pro": 84.3,
                "LiveCodeBenchV6": 77.1,
                "GPQA-Diamond": 82.7,
                "IFEval": 92.4
            },
            {
                "key": 13,
                "model": "MiniMax-M2",
                "org": "MiniMax",
                "num": "230B",
                "time": "2025/10/25",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 69.5,
                "HLE": 13.4,
                "AIME2025": 79.1,
                "MMLU-Pro": 81.6,
                "LiveCodeBenchV6": 74.0,
                "GPQA-Diamond": 78.7,
                "IFEval": 90.2
            },
            {
                "key": 14,
                "model": "JT-Think",
                "org": "China Mobile",
                "num": "N/A",
                "time": "2025/10/22",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 67.2,
                "HLE": 14.4,
                "AIME2025": 65.9,
                "MMLU-Pro": 85.6,
                "LiveCodeBenchV6": 64.9,
                "GPQA-Diamond": 77.7,
                "IFEval": 94.6
            },
            {
                "key": 15,
                "model": "GLM-4.6",
                "org": "Zhipu AI",
                "num": "355B",
                "time": "2025/9/30",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 73.3,
                "HLE": 19.3,
                "AIME2025": 90.3,
                "MMLU-Pro": 83.0,
                "LiveCodeBenchV6": 78.2,
                "GPQA-Diamond": 80.4,
                "IFEval": 88.7
            },
            {
                "key": 16,
                "model": "Qwen3-Next-80B-A3B-Thinking",
                "org": "Alibaba",
                "num": "80B",
                "time": "2025/9/12",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 69.5,
                "HLE": 13.5,
                "AIME2025": 89.0,
                "MMLU-Pro": 82.0,
                "LiveCodeBenchV6": 66.3,
                "GPQA-Diamond": 77.0,
                "IFEval": 89.5
            },
            {
                "key": 17,
                "model": "Qwen3-Next-80B-A3B-Instruct",
                "org": "Alibaba",
                "num": "80B",
                "time": "2025/9/12",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 62.5,
                "HLE": 8.0,
                "AIME2025": 69.2,
                "MMLU-Pro": 81.3,
                "LiveCodeBenchV6": 54.8,
                "GPQA-Diamond": 74.1,
                "IFEval": 87.6
            },
            {
                "key": 18,
                "model": "ERNIE-4.5-21B-A3B-Thinking",
                "org": "Baidu",
                "num": "21B",
                "time": "2025/9/9",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 57.0,
                "HLE": 6.5,
                "AIME2025": 76.2,
                "MMLU-Pro": 70.8,
                "LiveCodeBenchV6": 47.0,
                "GPQA-Diamond": 60.2,
                "IFEval": 81.2
            },
            {
                "key": 19,
                "model": "LongCat-Flash-Chat",
                "org": "Meituan",
                "num": "562B",
                "time": "2025/9/1",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 59.8,
                "HLE": 8.5,
                "AIME2025": 61.0,
                "MMLU-Pro": 81.0,
                "LiveCodeBenchV6": 48.2,
                "GPQA-Diamond": 70.1,
                "IFEval": 90.2
            },
            {
                "key": 20,
                "model": "Seed-OSS-36B-Instruct",
                "org": "ByteDance",
                "num": "36B",
                "time": "2025/8/20",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 66.5,
                "HLE": 9.5,
                "AIME2025": 86.0,
                "MMLU-Pro": 80.6,
                "LiveCodeBenchV6": 67.3,
                "GPQA-Diamond": 71.5,
                "IFEval": 83.7
            },
            {
                "key": 21,
                "model": "Qwen3-4B-Thinking-2507",
                "org": "Alibaba",
                "num": "4B",
                "time": "2025/8/7",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 60.6,
                "HLE": 6.0,
                "AIME2025": 80.0,
                "MMLU-Pro": 72.8,
                "LiveCodeBenchV6": 51.6,
                "GPQA-Diamond": 64.7,
                "IFEval": 88.5
            },
            {
                "key": 22,
                "model": "Qwen3-4B-Instruct-2507",
                "org": "Alibaba",
                "num": "4B",
                "time": "2025/8/7",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 47.2,
                "HLE": 5.1,
                "AIME2025": 46.9,
                "MMLU-Pro": 63.0,
                "LiveCodeBenchV6": 33.5,
                "GPQA-Diamond": 52.3,
                "IFEval": 82.4
            },
            {
                "key": 23,
                "model": "GPT-5-2025-08-07 (high)",
                "org": "OpenAI",
                "num": "N/A",
                "time": "2025/8/7",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 78.8,
                "HLE": 28.5,
                "AIME2025": 94.1,
                "MMLU-Pro": 86.2,
                "LiveCodeBenchV6": 83.2,
                "GPQA-Diamond": 86.4,
                "IFEval": 94.6
            },
            {
                "key": 24,
                "model": "GPT-5-mini-2025-08-07 (high)",
                "org": "OpenAI",
                "num": "N/A",
                "time": "2025/8/7",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 73.2,
                "HLE": 21.5,
                "AIME2025": 90.9,
                "MMLU-Pro": 70.2,
                "LiveCodeBenchV6": 80.8,
                "GPQA-Diamond": 82.7,
                "IFEval": 92.8
            },
            {
                "key": 25,
                "model": "GPT-5-nano-2025-08-07 (high)",
                "org": "OpenAI",
                "num": "N/A",
                "time": "2025/8/7",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 67.2,
                "HLE": 11.0,
                "AIME2025": 85.4,
                "MMLU-Pro": 75.0,
                "LiveCodeBenchV6": 74.0,
                "GPQA-Diamond": 67.8,
                "IFEval": 90.2
            },
            {
                "key": 26,
                "model": "gpt-oss-120b (high)",
                "org": "OpenAI",
                "num": "120B",
                "time": "2025/8/5",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 73.2,
                "HLE": 18.3,
                "AIME2025": 93.4,
                "MMLU-Pro": 79.7,
                "LiveCodeBenchV6": 78.4,
                "GPQA-Diamond": 78.9,
                "IFEval": 90.2
            },
            {
                "key": 27,
                "model": "gpt-oss-20b (high)",
                "org": "OpenAI",
                "num": "20B",
                "time": "2025/8/5",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 66.4,
                "HLE": 11.6,
                "AIME2025": 87.9,
                "MMLU-Pro": 72.8,
                "LiveCodeBenchV6": 68.4,
                "GPQA-Diamond": 68.9,
                "IFEval": 88.9
            },
            {
                "key": 28,
                "model": "Qwen3-30B-A3B-Thinking-2507",
                "org": "Alibaba",
                "num": "30B",
                "time": "2025/7/30",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 66.2,
                "HLE": 11.7,
                "AIME2025": 86.8,
                "MMLU-Pro": 79.5,
                "LiveCodeBenchV6": 59.4,
                "GPQA-Diamond": 70.1,
                "IFEval": 89.7
            },
            {
                "key": 29,
                "model": "Qwen3-30B-A3B-Instruct-2507",
                "org": "Alibaba",
                "num": "30B",
                "time": "2025/7/30",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 55.3,
                "HLE": 7.7,
                "AIME2025": 63.8,
                "MMLU-Pro": 73.9,
                "LiveCodeBenchV6": 38.7,
                "GPQA-Diamond": 63.6,
                "IFEval": 83.9
            },
            {
                "key": 30,
                "model": "GLM-4.5",
                "org": "Zhipu AI",
                "num": "358B",
                "time": "2025/7/29",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 69.2,
                "HLE": 16.9,
                "AIME2025": 85.8,
                "MMLU-Pro": 82.7,
                "LiveCodeBenchV6": 65.0,
                "GPQA-Diamond": 79.5,
                "IFEval": 85.4
            },
            {
                "key": 31,
                "model": "Qwen3-235B-A22B-Thinking-2507",
                "org": "Alibaba",
                "num": "235B",
                "time": "2025/7/25",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 71.8,
                "HLE": 18.5,
                "AIME2025": 90.9,
                "MMLU-Pro": 83.5,
                "LiveCodeBenchV6": 70.6,
                "GPQA-Diamond": 79.8,
                "IFEval": 87.8
            },
            {
                "key": 32,
                "model": "Qwen3-235B-A22B-Instruct-2507",
                "org": "Alibaba",
                "num": "235B",
                "time": "2025/7/22",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 61.3,
                "HLE": 12.3,
                "AIME2025": 69.5,
                "MMLU-Pro": 79.2,
                "LiveCodeBenchV6": 43.0,
                "GPQA-Diamond": 75.5,
                "IFEval": 88.3
            },
            {
                "key": 33,
                "model": "Kimi-K2-Instruct",
                "org": "Moonshot",
                "num": "1000B",
                "time": "2025/7/11",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 57.0,
                "HLE": 6.7,
                "AIME2025": 49.1,
                "MMLU-Pro": 78.4,
                "LiveCodeBenchV6": 46.8,
                "GPQA-Diamond": 72.7,
                "IFEval": 88.5
            },
            {
                "key": 34,
                "model": "ERNIE-4.5-300B-A47B-PT",
                "org": "Baidu",
                "num": "300B",
                "time": "2025/6/30",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 54.7,
                "HLE": 3.8,
                "AIME2025": 32.6,
                "MMLU-Pro": 78.7,
                "LiveCodeBenchV6": 41.0,
                "GPQA-Diamond": 83.0,
                "IFEval": 88.9
            },
            {
                "key": 35,
                "model": "Hunyuan-A13B-Instruct",
                "org": "Tencent",
                "num": "80B",
                "time": "2025/6/27",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 52.3,
                "HLE": 6.0,
                "AIME2025": 65.7,
                "MMLU-Pro": 73.8,
                "LiveCodeBenchV6": 43.9,
                "GPQA-Diamond": 64.1,
                "IFEval": 60.3
            },
            {
                "key": 36,
                "model": "Finix-P1-32B (Thinking)",
                "org": "Ant Group",
                "num": "N/A",
                "time": "2025/6/24",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 62.0,
                "HLE": 6.9,
                "AIME2025": 70.2,
                "MMLU-Pro": 82.7,
                "LiveCodeBenchV6": 54.9,
                "GPQA-Diamond": 68.7,
                "IFEval": 88.7
            },
            {
                "key": 37,
                "model": "Gemini-2.5-Pro",
                "org": "Google",
                "num": "N/A",
                "time": "2025/6/18",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 73.6,
                "HLE": 21.1,
                "AIME2025": 88.7,
                "MMLU-Pro": 85.8,
                "LiveCodeBenchV6": 71.3,
                "GPQA-Diamond": 84.7,
                "IFEval": 90.0
            },
            {
                "key": 38,
                "model": "doubao-seed-1.6-thinking-250615",
                "org": "ByteDance",
                "num": "N/A",
                "time": "2025/6/11",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 66.0,
                "HLE": 13.5,
                "AIME2025": 82.8,
                "MMLU-Pro": 83.1,
                "LiveCodeBenchV6": 61.8,
                "GPQA-Diamond": 75.9,
                "IFEval": 78.9
            },
            {
                "key": 39,
                "model": "DeepSeek-R1-0528",
                "org": "DeepSeek",
                "num": "671B",
                "time": "2025/5/28",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 68.1,
                "HLE": 14.4,
                "AIME2025": 89.0,
                "MMLU-Pro": 83.5,
                "LiveCodeBenchV6": 61.0,
                "GPQA-Diamond": 80.6,
                "IFEval": 80.0
            },
            {
                "key": 40,
                "model": "Claude Sonnet 4 (Thinking)",
                "org": "Anthropic",
                "num": "N/A",
                "time": "2025/5/22",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 61.8,
                "HLE": 8.7,
                "AIME2025": 68.7,
                "MMLU-Pro": 83.0,
                "LiveCodeBenchV6": 47.5,
                "GPQA-Diamond": 74.6,
                "IFEval": 88.3
            },
            {
                "key": 41,
                "model": "Qwen3-32B (Thinking)",
                "org": "Alibaba",
                "num": "32B",
                "time": "2025/4/29",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 61.3,
                "HLE": 8.5,
                "AIME2025": 70.3,
                "MMLU-Pro": 78.0,
                "LiveCodeBenchV6": 57.6,
                "GPQA-Diamond": 67.3,
                "IFEval": 86.0
            },
            {
                "key": 42,
                "model": "Qwen3-8B (Thinking)",
                "org": "Alibaba",
                "num": "8B",
                "time": "2025/4/29",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 56.5,
                "HLE": 5.5,
                "AIME2025": 66.2,
                "MMLU-Pro": 72.1,
                "LiveCodeBenchV6": 50.1,
                "GPQA-Diamond": 59.7,
                "IFEval": 85.6
            },
            {
                "key": 43,
                "model": "o3-2025-04-16",
                "org": "OpenAI",
                "num": "N/A",
                "time": "2025/4/16",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 72.8,
                "HLE": 21.3,
                "AIME2025": 85.1,
                "MMLU-Pro": 83.0,
                "LiveCodeBenchV6": 72.7,
                "GPQA-Diamond": 82.3,
                "IFEval": 92.2
            },
            {
                "key": 44,
                "model": "o4-mini-2025-04-16 (high)",
                "org": "OpenAI",
                "num": "N/A",
                "time": "2025/4/16",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 71.5,
                "HLE": 14.3,
                "AIME2025": 91.0,
                "MMLU-Pro": 82.4,
                "LiveCodeBenchV6": 71.0,
                "GPQA-Diamond": 77.4,
                "IFEval": 93.0
            },
            {
                "key": 45,
                "model": "GPT-4.1-20250414",
                "org": "OpenAI",
                "num": "N/A",
                "time": "2025/4/14",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 52.6,
                "HLE": 6.0,
                "AIME2025": 35.3,
                "MMLU-Pro": 81.0,
                "LiveCodeBenchV6": 37.2,
                "GPQA-Diamond": 68.1,
                "IFEval": 88.2
            },
            {
                "key": 46,
                "model": "Llama4-Maverick-17B-128E-Instruct",
                "org": "Meta",
                "num": "400B",
                "time": "2025/4/5",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 48.2,
                "HLE": 5.3,
                "AIME2025": 15.2,
                "MMLU-Pro": 79.6,
                "LiveCodeBenchV6": 33.2,
                "GPQA-Diamond": 69.1,
                "IFEval": 87.1
            },
            {
                "key": 47,
                "model": "Gemma-3-27B-it",
                "org": "Google",
                "num": "27B",
                "time": "2025/3/12",
                "update_time": "2026/3/9",
                "chat_or_base": "Chat",
                "Average": 42.1,
                "HLE": 4.2,
                "AIME2025": 22.4,
                "MMLU-Pro": 67.8,
                "LiveCodeBenchV6": 30.8,
                "GPQA-Diamond": 46.3,
                "IFEval": 81.0
            }
        ],
        "models": {
            "MiniCPM-SALA": {
                "desc": {
                    "zh-CN": "MiniCPM SALA是第一个有效整合稀疏和线性注意力的大规模混合模型，用于百万Token上下文建模。",
                    "en-US": "MiniCPM-SALA is the first large-scale hybrid model effectively integrating sparse and linear attention for million-token context modeling."
                },
                "org": "OpenBMB",
                "weight": [
                    "openbmb/MiniCPM-SALA",
                    "https://huggingface.co/openbmb/MiniCPM-SALA"
                ],
                "website-github": [
                    "OpenBMB",
                    "https://github.com/OpenBMB/MiniCPM/"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "9B",
                "time": "2026/2/11",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 1,
                "chat_or_base": "Chat"
            },
            "Qwen3.5-397B-A17B": {
                "desc": {
                    "zh-CN": "Qwen3.5 是阿里近期推出的重要基础模型，在多模态融合、混合高效架构、规模化强化学习、多语言覆盖及训练基建上实现全面升级，性能与实用性大幅提升。",
                    "en-US": "Qwen3.5 is a significant foundation model recently launched by Alibaba, featuring comprehensive upgrades in multimodal fusion, efficient hybrid architecture, scalable reinforcement learning, multilingual coverage and training infrastructure, with greatly improved performance and practicality."
                },
                "org": "Qwen",
                "weight": [
                    "Qwen/Qwen3.5-397B-A17B",
                    "https://huggingface.co/Qwen/Qwen3.5-397B-A17B"
                ],
                "website-github": [
                    "Qwen",
                    "https://github.com/QwenLM/Qwen3.5"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "397B",
                "time": "2026/2/16",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 2,
                "chat_or_base": "Chat"
            },
            "GLM-5": {
                "desc": {
                    "zh-CN": "GLM-5 是面向复杂系统工程与长时序智能体任务推出的新一代模型，在保持长上下文能力的同时大幅降低部署成本，延续了通过模型缩放提升 AGI 智能效率的核心路径",
                    "en-US": "GLM-5 is a new model launched for complex systems engineering and long-horizon agentic tasks, which significantly reduces deployment costs while maintaining long-context capabilities, and continues the core approach of improving AGI intelligence efficiency through model scaling."
                },
                "org": "Zhipu AI",
                "weight": [
                    "zai-org/GLM-5",
                    "https://huggingface.co/zai-org/GLM-5"
                ],
                "website-github": [
                    "Zhipu AI",
                    "https://z.ai/blog/glm-5"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "744B",
                "time": "2026/2/12",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 3,
                "chat_or_base": "Chat"
            },
            "MiniMax-M2.5": {
                "desc": {
                    "zh-CN": "MiniMax-M2.5是在数十万复杂真实环境中经大规模强化学习训练，在编程、智能体工具使用等多项高价值任务上达到 SOTA 水平，推理与任务分解效率高、执行速度显著提升，且成本极低、性价比突出，有望催生全新智能体应用的前沿模型。",
                    "en-US": "MiniMax-M2.5 is a cutting-edge model extensively trained via reinforcement learning in hundreds of thousands of complex real-world environments, achieving state-of-the-art performance in coding, agentic tool use, and many economically valuable tasks, with high reasoning efficiency, significantly improved execution speed, and extremely low cost, promising to enable innovative new agentic applications."
                },
                "org": "MiniMax",
                "weight": [
                    "MiniMaxAI/MiniMax-M2.5",
                    "https://huggingface.co/MiniMaxAI/MiniMax-M2.5"
                ],
                "website-github": [
                    "MiniMax",
                    "https://www.minimax.io/"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "230B",
                "time": "2026/2/12",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 4,
                "chat_or_base": "Chat"
            },
            "Kimi-K2.5": {
                "desc": {
                    "zh-CN": "由月之暗面公司最新推出的模型，具备更强代码能力、更擅长通用Agent任务。所使用版本号为kimi-k2.5。",
                    "en-US": "The latest model launched by Moonshot AI, featuring enhanced coding capabilities and superior performance in general Agent tasks, which adopts the version number kimi-k2.5."
                },
                "org": "Moonshot",
                "weight": [
                    "moonshotai/Kimi-K2.5",
                    ""
                ],
                "website-github": [
                    "Moonshot",
                    "https://www.moonshot.ai/"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "1000B",
                "time": "2026/1/27",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 5,
                "chat_or_base": "Chat"
            },
            "Step-3.5-Flash": {
                "desc": {
                    "zh-CN": "Step-3.5-Flash adopts the MoE architecture, achieving a balance between inference ability and operational efficiency, significantly improving the response speed of the model in complex agent workflows.",
                    "en-US": "Step-3.5-Flash采用了MoE架构，实现了推理能力与运行效率的平衡，显著提升了模型在复杂Agent工作流中的响应速度。"
                },
                "org": "StepFun",
                "weight": [
                    "stepfun-ai/Step-3.5-Flash",
                    "https://huggingface.co/stepfun-ai/Step-3.5-Flash"
                ],
                "website-github": [
                    "StepFun",
                    "https://static.stepfun.com/blog/step-3.5-flash/"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "196B",
                "time": "2026/2/2",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 6,
                "chat_or_base": "Chat"
            },
            "TeleChat-thinking": {
                "desc": {
                    "zh-CN": "中国电信人工智能研究院（TeleAI）推出的大语言模型。",
                    "en-US": "The large language model launched by TeleAI."
                },
                "org": "TeleAI",
                "weight": [
                    "",
                    ""
                ],
                "website-github": [
                    "TeleAI",
                    "https://www.teleai.com.cn/"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "N/A",
                "time": "2025/12/29",
                "update_time": "2026/3/9",
                "release": "API",
                "rank": 7,
                "chat_or_base": "Chat"
            },
            "GLM-4.7": {
                "desc": {
                    "zh-CN": "GLM发布的新一代GLM-4.7系列开源推理模型。",
                    "en-US": "The new GLM-4.7 reasoning model series released by GLM."
                },
                "org": "Zhipu AI",
                "weight": [
                    "zai-org/GLM-4.7",
                    "https://huggingface.co/zai-org/GLM-4.7"
                ],
                "website-github": [
                    "Zhipu AI",
                    "https://z.ai/model-api"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "355B",
                "time": "2025/12/23",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 8,
                "chat_or_base": "Chat"
            },
            "MiMo-V2-Flash": {
                "desc": {
                    "zh-CN": "MiMo-V2-Flash is an MoE language model, designed for high-speed reasoning and agentic workflows, it utilizes a novel hybrid attention architecture and MTP to achieve state-of-the-art performance while significantly reducing inference costs.",
                    "en-US": "MiMo-V2-Flash是一个MoE语言模型，专为高速推理和代理工作流程而设计，利用新颖的混合注意力架构和MTP来实现最先进的性能，同时显著降低推理成本。"
                },
                "org": "Xiaomi",
                "weight": [
                    "XiaomiMiMo/MiMo-V2-Flash",
                    "https://huggingface.co/XiaomiMiMo/MiMo-V2-Flash"
                ],
                "website-github": [
                    "Xiaomi",
                    "https://platform.xiaomimimo.com/#/docs/welcome"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "309B",
                "time": "2025/12/17",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 9,
                "chat_or_base": "Chat"
            },
            "DeepSeek-V3.2": {
                "desc": {
                    "zh-CN": "DeepSeek最新推出的V3.2模型正式版，该模型在保持卓越推理能力和智能体性能的同时，实现了高计算效率的平衡。",
                    "en-US": "The official release of DeepSeek’s newly launched V3.2 model has just been announced. While maintaining outstanding reasoning capabilities and agent performance, the model also strikes a balance with high computational efficiency."
                },
                "org": "DeepSeek",
                "weight": [
                    "deepseek-ai/DeepSeek-V3.2",
                    "https://huggingface.co/deepseek-ai/DeepSeek-V3.2"
                ],
                "website-github": [
                    "DeepSeek",
                    "https://huggingface.co/deepseek-ai/DeepSeek-V3.2"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "671B",
                "time": "2025/12/1",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 10,
                "chat_or_base": "Chat"
            },
            "DeepSeek-V3.2-Speciale": {
                "desc": {
                    "zh-CN": "DeepSeekV3.2的高计算版本，具有更强的计算和推理能力。",
                    "en-US": "The high-compute version of DeepSeekV3.2, offering stronger computational and reasoning capabilities."
                },
                "org": "DeepSeek",
                "weight": [
                    "deepseek-ai/DeepSeek-V3.2-Speciale",
                    "https://huggingface.co/deepseek-ai/DeepSeek-V3.2-Speciale"
                ],
                "website-github": [
                    "DeepSeek",
                    "https://huggingface.co/deepseek-ai/DeepSeek-V3.2-Speciale"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "671B",
                "time": "2025/12/1",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 11,
                "chat_or_base": "Chat"
            },
            "Gemini-3-Pro-Preview": {
                "desc": {
                    "zh-CN": "gemini3-pro模型的预览版本。",
                    "en-US": "The preview version of the Gemini 3 pro."
                },
                "org": "Google",
                "weight": [
                    "",
                    ""
                ],
                "website-github": [
                    "Google",
                    "https://developers.googleblog.com/new-gemini-api-updates-for-gemini-3/"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "N/A",
                "time": "2025/11/19",
                "update_time": "2026/3/9",
                "release": "API",
                "rank": 12,
                "chat_or_base": "Chat"
            },
            "Kimi-K2-Thinking": {
                "desc": {
                    "zh-CN": "由月之暗面公司最新推出的深度思考模型，具备更强代码能力、更擅长通用Agent任务的MoE架构模型。",
                    "en-US": "The latest deep reasoning model launched by Moonshot AI, featuring enhanced coding capabilities and superior performance in general Agent tasks, built on an MoE architecture."
                },
                "org": "Moonshot",
                "weight": [
                    "moonshotai/Kimi-K2-Thinking",
                    "https://huggingface.co/moonshotai/Kimi-K2-Thinking"
                ],
                "website-github": [
                    "Moonshot",
                    "https://moonshotai.github.io/Kimi-K2/thinking.html"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "1000B",
                "time": "2025/11/6",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 13,
                "chat_or_base": "Chat"
            },
            "MiniMax-M2": {
                "desc": {
                    "zh-CN": "Minimax最新发布的旗舰模型，具有卓越的通用能力、先进的编程能力、强大的智能体性能以及高效的架构设计。",
                    "en-US": "The latest flagship model released by Minimax features excellent general capabilities, advanced programming capabilities, powerful agent performance, and efficient architecture design."
                },
                "org": "MiniMax",
                "weight": [
                    "MiniMaxAI/MiniMax-M2",
                    "https://huggingface.co/MiniMaxAI/MiniMax-M2"
                ],
                "website-github": [
                    "MiniMax",
                    "https://www.minimaxi.com/platform"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "230B",
                "time": "2025/10/25",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 14,
                "chat_or_base": "Chat"
            },
            "JT-Think": {
                "desc": {
                    "zh-CN": "中国移动推出的闭源深度思考模型。",
                    "en-US": "A close-source deep reasoning model launched by China Mobile."
                },
                "org": "China Mobile",
                "weight": [
                    "",
                    ""
                ],
                "website-github": [
                    "China Mobile",
                    "https://jiutian.10086.cn/portal/#/home"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "N/A",
                "time": "2025/10/22",
                "update_time": "2026/3/9",
                "release": "API",
                "rank": 15,
                "chat_or_base": "Chat"
            },
            "GLM-4.6": {
                "desc": {
                    "zh-CN": "GLM发布的新一代GLM-4.6系列开源推理模型。",
                    "en-US": "The new GLM-4.6 reasoning model series released by GLM."
                },
                "org": "Zhipu AI",
                "weight": [
                    "zai-org/GLM-4.6",
                    "https://huggingface.co/zai-org/GLM-4.6"
                ],
                "website-github": [
                    "Zhipu AI",
                    "https://z.ai/blog/glm-4.6"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "355B",
                "time": "2025/9/30",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 16,
                "chat_or_base": "Chat"
            },
            "Qwen3-Next-80B-A3B-Thinking": {
                "desc": {
                    "zh-CN": "Qwen模型中最新发布的Qwen3-Next系列，通过创新的模型架构改进Scaling效率。",
                    "en-US": "The latest released Qwen3-Next series in Qwen models, improving scaling efficiency through innovative model architecture."
                },
                "org": "Alibaba",
                "weight": [
                    "Qwen/Qwen3-Next-80B-A3B-Thinking",
                    "https://huggingface.co/Qwen/Qwen3-Next-80B-A3B-Thinking"
                ],
                "website-github": [
                    "Qwen3",
                    "https://github.com/QwenLM/Qwen3"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "80B",
                "time": "2025/9/12",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 17,
                "chat_or_base": "Chat"
            },
            "Qwen3-Next-80B-A3B-Instruct": {
                "desc": {
                    "zh-CN": "Qwen模型中最新发布的Qwen3-Next系列，通过创新的模型架构改进Scaling效率。",
                    "en-US": "The latest released Qwen3-Next series in Qwen models, improving scaling efficiency through innovative model architecture."
                },
                "org": "Alibaba",
                "weight": [
                    "Qwen/Qwen3-Next-80B-A3B-Instruct",
                    "https://huggingface.co/Qwen/Qwen3-Next-80B-A3B-Instruct"
                ],
                "website-github": [
                    "Qwen3",
                    "https://github.com/QwenLM/Qwen3"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "80B",
                "time": "2025/9/12",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 18,
                "chat_or_base": "Chat"
            },
            "ERNIE-4.5-21B-A3B-Thinking": {
                "desc": {
                    "zh-CN": "百度在文心一言4.5系列的基础上发布的新模型，其思维能力得到扩展，推理质量和深度得到提升。",
                    "en-US": "The new model released by Baidu on the basis of ERNIE-4.5 series, which has expanded its thinking ability and improved the quality and depth of reasoning."
                },
                "org": "Baidu",
                "weight": [
                    "baidu/ERNIE-4.5-21B-A3B-Thinking",
                    "https://huggingface.co/baidu/ERNIE-4.5-21B-A3B-Thinking"
                ],
                "website-github": [
                    "Baidu",
                    "https://yiyan.baidu.com/blog/posts/ernie4.5"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "21B",
                "time": "2025/9/9",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 19,
                "chat_or_base": "Chat"
            },
            "LongCat-Flash-Chat": {
                "desc": {
                    "zh-CN": "美团推出的一款名为 LongCat-Flash 的非推理开源大模型，以极高的推理效率著称。",
                    "en-US": "An open-sourced non-reasoning LLM named LongCat-Flash launched by Meituan, known for its extremely high inference efficiency."
                },
                "org": "Meituan",
                "weight": [
                    "meituan-longcat/LongCat-Flash-Chat",
                    "https://huggingface.co/meituan-longcat/LongCat-Flash-Chat"
                ],
                "website-github": [
                    "LongCat",
                    "https://longcat.chat/"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "562B",
                "time": "2025/9/1",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 20,
                "chat_or_base": "Chat"
            },
            "Seed-OSS-36B-Instruct": {
                "desc": {
                    "zh-CN": "Seed-OSS是字节跳动开发的一系列开源大型语言模型，旨在提供强大的长上下文、推理、代理和通用功能，以及通用的开发人员友好功能。尽管仅使用12T令牌进行训练，但Seed-OSS在几个流行的开放基准测试中取得了出色的性能。",
                    "en-US": "Seed-OSS is a series of open-source large language models developed by ByteDance's Seed Team, designed for powerful long-context, reasoning, agent and general capabilities, and versatile developer-friendly features. Although trained with only 12T tokens, Seed-OSS achieves excellent performance on several popular open benchmarks."
                },
                "org": "ByteDance",
                "weight": [
                    "ByteDance-Seed/Seed-OSS-36B-Instruct",
                    "https://huggingface.co/ByteDance-Seed/Seed-OSS-36B-Instruct"
                ],
                "website-github": [
                    "ByteDance-Seed",
                    "https://github.com/ByteDance-Seed/seed-oss"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "36B",
                "time": "2025/8/20",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 21,
                "chat_or_base": "Chat"
            },
            "Qwen3-4B-Thinking-2507": {
                "desc": {
                    "zh-CN": "Qwen3是Qwen系列中最新一代的大型语言模型，提供了一套全面的密集模型和MoE模型。",
                    "en-US": "Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. "
                },
                "org": "Alibaba",
                "weight": [
                    "Qwen/Qwen3-4B-Thinking-2507",
                    "https://huggingface.co/Qwen/Qwen3-4B-Thinking-2507"
                ],
                "website-github": [
                    "Qwen3",
                    "https://github.com/QwenLM/Qwen3"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "4B",
                "time": "2025/8/7",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 22,
                "chat_or_base": "Chat"
            },
            "Qwen3-4B-Instruct-2507": {
                "desc": {
                    "zh-CN": "Qwen3是Qwen系列中最新一代的大型语言模型，提供了一套全面的密集模型和MoE模型。",
                    "en-US": "Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. "
                },
                "org": "Alibaba",
                "weight": [
                    "Qwen/Qwen3-4B-Instruct-2507",
                    "https://huggingface.co/Qwen/Qwen3-4B-Instruct-2507"
                ],
                "website-github": [
                    "Qwen3",
                    "https://github.com/QwenLM/Qwen3"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "4B",
                "time": "2025/8/7",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 23,
                "chat_or_base": "Chat"
            },
            "GPT-5-2025-08-07 (high)": {
                "desc": {
                    "zh-CN": "OpenAI官方最新推出的备受外界期待的、性能更强的新一代人工智能模型GPT-5。",
                    "en-US": "The latest, highly anticipated and powerful next-generation LLM named GPT-5 that officially launched by OpenAI."
                },
                "org": "OpenAI",
                "weight": [
                    "",
                    ""
                ],
                "website-github": [
                    "OpenAI",
                    "https://openai.com/index/introducing-gpt-5/"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "N/A",
                "time": "2025/8/7",
                "update_time": "2026/3/9",
                "release": "API",
                "rank": 24,
                "chat_or_base": "Chat"
            },
            "GPT-5-mini-2025-08-07 (high)": {
                "desc": {
                    "zh-CN": "OpenAI官方最新推出的备受外界期待的、性能更强的新一代人工智能模型GPT-5。",
                    "en-US": "The latest, highly anticipated and powerful next-generation LLM named GPT-5 that officially launched by OpenAI."
                },
                "org": "OpenAI",
                "weight": [
                    "",
                    ""
                ],
                "website-github": [
                    "OpenAI",
                    "https://openai.com/index/introducing-gpt-5/"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "N/A",
                "time": "2025/8/7",
                "update_time": "2026/3/9",
                "release": "API",
                "rank": 25,
                "chat_or_base": "Chat"
            },
            "GPT-5-nano-2025-08-07 (high)": {
                "desc": {
                    "zh-CN": "OpenAI官方最新推出的备受外界期待的、性能更强的新一代人工智能模型GPT-5。",
                    "en-US": "The latest, highly anticipated and powerful next-generation LLM named GPT-5 that officially launched by OpenAI."
                },
                "org": "OpenAI",
                "weight": [
                    "",
                    ""
                ],
                "website-github": [
                    "OpenAI",
                    "https://openai.com/index/introducing-gpt-5/"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "N/A",
                "time": "2025/8/7",
                "update_time": "2026/3/9",
                "release": "API",
                "rank": 26,
                "chat_or_base": "Chat"
            },
            "gpt-oss-120b (high)": {
                "desc": {
                    "zh-CN": "由OpenAI发布的近年来首个开源模型系列gpt-oss。",
                    "en-US": "The GPT-OSS series, the first open-source model family released by OpenAI in recent years."
                },
                "org": "OpenAI",
                "weight": [
                    "openai/gpt-oss-120b",
                    "https://huggingface.co/openai/gpt-oss-120b"
                ],
                "website-github": [
                    "OpenAI",
                    "https://openai.com/index/introducing-gpt-oss/"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "120B",
                "time": "2025/8/5",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 27,
                "chat_or_base": "Chat"
            },
            "gpt-oss-20b (high)": {
                "desc": {
                    "zh-CN": "由OpenAI发布的近年来首个开源模型系列gpt-oss。",
                    "en-US": "The GPT-OSS series, the first open-source model family released by OpenAI in recent years."
                },
                "org": "OpenAI",
                "weight": [
                    "openai/gpt-oss-20b",
                    "https://huggingface.co/openai/gpt-oss-20b"
                ],
                "website-github": [
                    "OpenAI",
                    "https://openai.com/index/introducing-gpt-oss/"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "20B",
                "time": "2025/8/5",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 28,
                "chat_or_base": "Chat"
            },
            "Qwen3-30B-A3B-Thinking-2507": {
                "desc": {
                    "zh-CN": "Qwen3是Qwen系列中最新一代的大型语言模型，提供了一套全面的密集模型和MoE模型。",
                    "en-US": "Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. "
                },
                "org": "Alibaba",
                "weight": [
                    "Qwen/Qwen3-30B-A3B-Thinking-2507",
                    "https://huggingface.co/Qwen/Qwen3-30B-A3B-Thinking-2507"
                ],
                "website-github": [
                    "Qwen3",
                    "https://github.com/QwenLM/Qwen3"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "30B",
                "time": "2025/7/30",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 29,
                "chat_or_base": "Chat"
            },
            "Qwen3-30B-A3B-Instruct-2507": {
                "desc": {
                    "zh-CN": "Qwen3是Qwen系列中最新一代的大型语言模型，提供了一套全面的密集模型和MoE模型。",
                    "en-US": "Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. "
                },
                "org": "Alibaba",
                "weight": [
                    "Qwen/Qwen3-30B-A3B-Instruct-2507",
                    "https://huggingface.co/Qwen/Qwen3-30B-A3B-Instruct-2507"
                ],
                "website-github": [
                    "Qwen3",
                    "https://github.com/QwenLM/Qwen3"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "30B",
                "time": "2025/7/30",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 30,
                "chat_or_base": "Chat"
            },
            "GLM-4.5": {
                "desc": {
                    "zh-CN": "GLM发布的新一代GLM-4.5系列开源推理模型。",
                    "en-US": "The new GLM-4.5 reasoning model series released by GLM."
                },
                "org": "Zhipu AI",
                "weight": [
                    "zai-org/GLM-4.5",
                    "https://huggingface.co/zai-org/GLM-4.5"
                ],
                "website-github": [
                    "GLM-4.5",
                    "https://z.ai/blog/glm-4.5"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "358B",
                "time": "2025/7/29",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 31,
                "chat_or_base": "Chat"
            },
            "Qwen3-235B-A22B-Thinking-2507": {
                "desc": {
                    "zh-CN": "Qwen3是Qwen系列中最新一代的大型语言模型，提供了一套全面的密集模型和MoE模型。",
                    "en-US": "Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. "
                },
                "org": "Alibaba",
                "weight": [
                    "Qwen/Qwen3-235B-A22B-Thinking-2507",
                    "https://huggingface.co/Qwen/Qwen3-235B-A22B-Thinking-2507"
                ],
                "website-github": [
                    "Qwen3",
                    "https://github.com/QwenLM/Qwen3"
                ],
                "article": [
                    "report",
                    "https://qwenlm.github.io/zh/blog/qwen3/"
                ],
                "num": "235B",
                "time": "2025/7/25",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 32,
                "chat_or_base": "Chat"
            },
            "Qwen3-235B-A22B-Instruct-2507": {
                "desc": {
                    "zh-CN": "Qwen3是Qwen系列中最新一代的大型语言模型，提供了一套全面的密集模型和MoE模型。",
                    "en-US": "Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. "
                },
                "org": "Alibaba",
                "weight": [
                    "Qwen/Qwen3-235B-A22B-Instruct-2507",
                    "https://huggingface.co/Qwen/Qwen3-235B-A22B-Instruct-2507"
                ],
                "website-github": [
                    "Qwen3",
                    "https://github.com/QwenLM/Qwen3"
                ],
                "article": [
                    "report",
                    "https://qwenlm.github.io/zh/blog/qwen3/"
                ],
                "num": "235B",
                "time": "2025/7/22",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 33,
                "chat_or_base": "Chat"
            },
            "Kimi-K2-Instruct": {
                "desc": {
                    "zh-CN": "由月之暗面公司最新推出的模型，具备更强代码能力、更擅长通用Agent任务的MoE架构模型。",
                    "en-US": "The latest model launched by Moonshot AI, featuring enhanced coding capabilities and superior performance in general Agent tasks, built on an MoE architecture."
                },
                "org": "Moonshot",
                "weight": [
                    "moonshotai/Kimi-K2-Instruct",
                    "https://huggingface.co/moonshotai/Kimi-K2-Instruct"
                ],
                "website-github": [
                    "Moonshot",
                    "https://moonshotai.github.io/Kimi-K2/"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "1000B",
                "time": "2025/7/11",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 34,
                "chat_or_base": "Chat"
            },
            "ERNIE-4.5-300B-A47B-PT": {
                "desc": {
                    "zh-CN": "百度文心一言系列模型的最新开源版本。",
                    "en-US": "The latest open source version of Baidu ERNIE series LLMs."
                },
                "org": "Baidu",
                "weight": [
                    "baidu/ERNIE-4.5-300B-A47B-PT",
                    "https://huggingface.co/baidu/ERNIE-4.5-300B-A47B-PT"
                ],
                "website-github": [
                    "Baidu",
                    "https://yiyan.baidu.com/blog/posts/ernie4.5"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "300B",
                "time": "2025/6/30",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 35,
                "chat_or_base": "Chat"
            },
            "Hunyuan-A13B-Instruct": {
                "desc": {
                    "zh-CN": "腾讯最新发布的MOE模型。",
                    "en-US": "The latest MOE model released by Tencent."
                },
                "org": "Tencent",
                "weight": [
                    "tencent/Hunyuan-A13B-Instruct",
                    "https://huggingface.co/tencent/Hunyuan-A13B-Instruct"
                ],
                "website-github": [
                    "Tencent",
                    "https://hunyuan.tencent.com/"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "80B",
                "time": "2025/6/27",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 36,
                "chat_or_base": "Chat"
            },
            "Finix-P1-32B (Thinking)": {
                "desc": {
                    "zh-CN": "由蚂蚁集团发布的金融领域大模型。",
                    "en-US": "The financial domain LLM released by Ant Group"
                },
                "org": "Ant Group",
                "weight": [
                    "",
                    ""
                ],
                "website-github": [
                    "Ant Group",
                    "https://antfinix.alipay.com/"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "N/A",
                "time": "2025/6/24",
                "update_time": "2026/3/9",
                "release": "API",
                "rank": 37,
                "chat_or_base": "Chat"
            },
            "Gemini-2.5-Pro": {
                "desc": {
                    "zh-CN": "gemini2.5-pro模型的正式发布版本。",
                    "en-US": "The official release version of the Gemini 2.5 pro."
                },
                "org": "Google",
                "weight": [
                    "",
                    ""
                ],
                "website-github": [
                    "Google",
                    "https://developers.googleblog.com/en/gemini-2-5-thinking-model-updates/"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "N/A",
                "time": "2025/6/18",
                "update_time": "2026/3/9",
                "release": "API",
                "rank": 38,
                "chat_or_base": "Chat"
            },
            "doubao-seed-1.6-thinking-250615": {
                "desc": {
                    "zh-CN": "字节豆包推出的seed系列模型的最新版，支持思考模式。",
                    "en-US": "The latest version of the seed series model launched by ByteDance, whitch supports the thinking mode."
                },
                "org": "ByteDance",
                "weight": [
                    "",
                    ""
                ],
                "website-github": [
                    "Doubao",
                    "https://console.volcengine.com/ark"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "N/A",
                "time": "2025/6/11",
                "update_time": "2026/3/9",
                "release": "API",
                "rank": 39,
                "chat_or_base": "Chat"
            },
            "DeepSeek-R1-0528": {
                "desc": {
                    "zh-CN": "Deepseek R1的最新版本。",
                    "en-US": "The latest version of Deepseek R1."
                },
                "org": "DeepSeek",
                "weight": [
                    "deepseek-ai/DeepSeek-R1-0528",
                    "https://huggingface.co/deepseek-ai/DeepSeek-R1-0528"
                ],
                "website-github": [
                    "DeepSeek",
                    "https://github.com/deepseek-ai/DeepSeek-R1"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "671B",
                "time": "2025/5/28",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 40,
                "chat_or_base": "Chat"
            },
            "Claude Sonnet 4 (Thinking)": {
                "desc": {
                    "zh-CN": "Anthropic最新发布的混合推理模型系列，支持更好的代码编写、工件调用等功能。",
                    "en-US": "The latest hybrid inference model series released by Anthropic, which support better coding, function calling, and other features."
                },
                "org": "Anthropic",
                "weight": [
                    "",
                    ""
                ],
                "website-github": [
                    "Anthropic",
                    "https://www.anthropic.com/claude/sonnet"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "N/A",
                "time": "2025/5/22",
                "update_time": "2026/3/9",
                "release": "API",
                "rank": 41,
                "chat_or_base": "Chat"
            },
            "Qwen3-32B (Thinking)": {
                "desc": {
                    "zh-CN": "Qwen3是Qwen系列中最新一代的大型语言模型，提供了一套全面的密集模型和MoE模型。",
                    "en-US": "Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. "
                },
                "org": "Alibaba",
                "weight": [
                    "Qwen/Qwen3-32B",
                    "https://huggingface.co/Qwen/Qwen3-32B"
                ],
                "website-github": [
                    "Qwen3",
                    "https://github.com/QwenLM/Qwen3"
                ],
                "article": [
                    "report",
                    "https://qwenlm.github.io/zh/blog/qwen3/"
                ],
                "num": "32B",
                "time": "2025/4/29",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 42,
                "chat_or_base": "Chat"
            },
            "Qwen3-8B (Thinking)": {
                "desc": {
                    "zh-CN": "Qwen3是Qwen系列中最新一代的大型语言模型，提供了一套全面的密集模型和MoE模型。",
                    "en-US": "Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. "
                },
                "org": "Alibaba",
                "weight": [
                    "Qwen/Qwen3-8B",
                    "https://huggingface.co/Qwen/Qwen3-8B"
                ],
                "website-github": [
                    "Qwen3",
                    "https://github.com/QwenLM/Qwen3"
                ],
                "article": [
                    "report",
                    "https://qwenlm.github.io/zh/blog/qwen3/"
                ],
                "num": "8B",
                "time": "2025/4/29",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 43,
                "chat_or_base": "Chat"
            },
            "o3-2025-04-16": {
                "desc": {
                    "zh-CN": "o3是openai发布的最新深度思考模型，其基线指标大幅超越上一代模型。",
                    "en-US": "O3 is the latest reasoning model released by OpenAI, with baseline metrics significantly surpassing previous models."
                },
                "org": "OpenAI",
                "weight": [
                    "",
                    ""
                ],
                "website-github": [
                    "OpenAI",
                    "https://openai.com/index/introducing-o3-and-o4-mini/"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "N/A",
                "time": "2025/4/16",
                "update_time": "2026/3/9",
                "release": "API",
                "rank": 44,
                "chat_or_base": "Chat"
            },
            "o4-mini-2025-04-16 (high)": {
                "desc": {
                    "zh-CN": "high模式下的o4-mini，响应时间更长，具有更强的推理能力。",
                    "en-US": "o4-mini in high mode, which has a longer response time and stronger reasoning ability."
                },
                "org": "OpenAI",
                "weight": [
                    "",
                    ""
                ],
                "website-github": [
                    "OpenAI",
                    "https://openai.com/index/introducing-o3-and-o4-mini/"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "N/A",
                "time": "2025/4/16",
                "update_time": "2026/3/9",
                "release": "API",
                "rank": 45,
                "chat_or_base": "Chat"
            },
            "GPT-4.1-20250414": {
                "desc": {
                    "zh-CN": "由OpenAI推出的GPT-4.1系列模型，在编码、指令遵循和长上下文方面有重大改进。",
                    "en-US": "The GPT-4.1 series models launched by OpenAI, which have significant improvements in encoding, instruction adherence, and long context."
                },
                "org": "OpenAI",
                "weight": [
                    "",
                    ""
                ],
                "website-github": [
                    "OpenAI",
                    "https://platform.openai.com/docs/models/gpt-4.1"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "N/A",
                "time": "2025/4/14",
                "update_time": "2026/3/9",
                "release": "API",
                "rank": 46,
                "chat_or_base": "Chat"
            },
            "Llama4-Maverick-17B-128E-Instruct": {
                "desc": {
                    "zh-CN": "Llama 4模型是自回归语言模型，它使用混合专家（MoE）架构，并结合了原生多模态的早期融合",
                    "en-US": "The Llama 4 models are auto-regressive language models that use a mixture-of-experts (MoE) architecture and incorporate early fusion for native multimodality"
                },
                "org": "Meta",
                "weight": [
                    "meta-llama/Llama-4-Maverick-17B-128E-Instruct",
                    "https://huggingface.co/meta-llama/Llama-4-Maverick-17B-128E-Instruct"
                ],
                "website-github": [
                    "Llama4-Maverick",
                    "https://www.llama.com/"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "400B",
                "time": "2025/4/5",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 47,
                "chat_or_base": "Chat"
            },
            "Gemma-3-27B-it": {
                "desc": {
                    "zh-CN": "Gemma 是来自谷歌的一系列轻量级、最先进的开源模型家族，这些模型基于创建Gemini模型相同的研究和技术构建而成。",
                    "en-US": "Gemma is a family of lightweight, state-of-the-art open models from Google, built from the same research and technology used to create the Gemini models."
                },
                "org": "Google",
                "weight": [
                    "google/gemma-3-27b-it",
                    "https://huggingface.co/google/gemma-3-27b-it"
                ],
                "website-github": [
                    "Google",
                    "https://ai.google.dev/gemma/docs"
                ],
                "article": [
                    "",
                    ""
                ],
                "num": "27B",
                "time": "2025/3/12",
                "update_time": "2026/3/9",
                "release": "OpenSource",
                "rank": 48,
                "chat_or_base": "Chat"
            }
        },
        "update_time": "2026/3/9"
    }
}