大模型 Agent 能力评测排行榜
本页面提供大模型 Agent 能力评测排行榜，涵盖 Terminal-Bench 4.0、τ²-Bench、MCP-Atlas、Tool Decathlon、OSWorld-Verified 等主流 Agent 评测基准，深度对比 GPT、Claude、Qwen、DeepSeek 等模型的工具使用、任务规划与自主执行能力。
数据更新于 2026-10-01 08:16:28
截至 2026年10月，本页覆盖 Terminal-Bench 4.0, τ²-Bench, MCP-Atlas, Tool Decathlon 等评测基准，聚焦 大模型 Agent 能力评测排行榜 方向的模型对比。
点击模型名称可进入详情页查看上下文长度、许可方式与 API 价格。数据口径说明见 数据方法论。
榜单亮点
按 Terminal-Bench 4.0 排序
大模型性能评测结果
数据来源：DataLearnerAI点击任意行查看模型详情；勾选左侧 可对比最多 4 个模型。
Terminal-Bench 4.058.18
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.057.88
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.057.88
τ²-Bench—
MCP-Atlas87.20
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.057.88
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.057.88
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.057.40
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.054.55
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.054.24
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.053.94
τ²-Bench—
MCP-Atlas85.80
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.053.94
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.051.82
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.050.61
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.050.30
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.044.85
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.044.55
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.043.94
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.043.33
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.041.82
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
有条件商用
Terminal-Bench 4.037.58
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.037.27
τ²-Bench—
MCP-Atlas81.80
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.034.85
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.034.85
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified82.00
免费商用
Terminal-Bench 4.033.30
τ²-Bench—
MCP-Atlas85.60
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.033.30
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.032.80
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
免费商用
Terminal-Bench 4.028.80
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified80.80
免费商用
Terminal-Bench 4.026.80
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
免费商用
Terminal-Bench 4.025.30
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
有条件商用
Terminal-Bench 4.024.70
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.023.64
τ²-Bench—
MCP-Atlas82.20
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.021.52
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.020.70
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.019.70
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.019.70
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.019.09
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.018.70
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
有条件商用
Terminal-Bench 4.017.27
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.017.20
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.016.70
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.014.60
τ²-Bench—
MCP-Atlas75.30
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.014.60
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
免费商用
Terminal-Bench 4.014.60
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
Terminal-Bench 4.013.60
τ²-Bench—
MCP-Atlas—
Tool Decathlon—
OSWorld-Verified—
闭源
排序：
已显示 50 / 178 个模型查看 Terminal-Bench 4.0 基准测试完整页面