τ²-Bench
为了解决大模型的Agent操作依赖交互和人工处理这个问题,普林斯顿大学与 Sierra Research 的研究团队在 2025 年 6 月提出了 τ²-Bench(Tau-Squared Benchmark),并发布了论文《τ²-Bench: Evaluating Conversational Agents in a Dual-Control Environment》。 它是对早期 τ-Bench 的扩展版本,旨在建立一种标准化方法,评估智能体在与用户共同作用于环境时的表现。
更新于 2026-04-08
2,021 次浏览
问题数量
2285
发布机构
个人
评测类别
Agent能力评测
评测指标
Accuracy
支持语言
中文
难度等级
中等难度
简介
用于评估大模型智能体在双控(人机协同)环境下任务执行与沟通能力的开放基准
相关资源
τ²-Bench Model Score Leaderboard
Source: DataLearnerAI
Data sourced primarily from official releases (GitHub, Hugging Face, papers), then benchmark leaderboards, then third-party evaluators. Learn about our data methodology
模式说明:
normal
thinking
low
medium
high
deeper thinking
parallel_thinking
图表加载中...
Latest τ²-Bench model rankings and full benchmark leaderboard
Browse the latest scores, model modes, release dates, and parameter sizes for τ²-Bench.
τ²-Bench详细排名数据表格
| 排名 | 模型 | |||
|---|---|---|---|---|
| 1 | Claude Opus 4.6 Deep ThinkingTools | 91.89 | 2026-02-05 | 未知 |
| 2 | Gemini 3.1 Pro Preview Thinking Level · HighTools | 90.8 | 2026-02-20 | 未知 |
| 3 | Gemini 3.0 Flash Thinking Level · MediumTools | 90.2 | 2025-12-17 | 未知 |
| 4 | GLM-5 Thinking Level · MediumTools | 89.7 | 2026-02-11 | 7440 |
| 5 | Step 3.5 Flash Thinking Level · MediumTools | 88.2 | 2026-02-02 | 1960 |
| 6 | GLM-4.7 Thinking Level · MediumTools | 87.4 | 2025-12-22 | 3580 |
| 7 | Qwen3.5-397B-A17B Thinking Level · MediumTools | 86.7 | 2026-02-16 | 397 |
| 8 | Gemini 3.0 Pro (Preview 11-2025) Thinking Level · MediumTools | 85.4 | 2025-11-18 | 未知 |
| 9 | Claude Sonnet 4.5 Thinking Level · MediumTools | 84.7 | 2025-09-30 | 未知 |
| 10 | Grok 4.1 Fast Thinking Level · MediumTools | 82.71 | 2025-11-19 | 未知 |
| 11 | Qwen3-Max-Thinking Thinking Level · MediumTools | 82.1 | 2026-01-26 | 10000 |
| 12 | GPT-5.2 Deep ThinkingTools | 82 | 2025-12-11 | 未知 |
| 13 | Claude Opus 4.5 Thinking Level · MediumTools | 81.99 | 2025-11-25 | 未知 |
| 14 | DeepSeek V3.2 Thinking Level · MediumTools | 80.3 | 2025-12-01 | 6710 |
| 15 | GPT-5 Thinking Level · MediumTools | 80 | 2025-08-07 | 未知 |
| 16 | GLM-4.7-Flash Thinking Level · MediumTools | 79.5 | 2026-01-19 | 310 |
| 17 | Qwen3.5-27B Thinking Level · MediumTools | 79 | 2026-02-25 | 270 |
| 18 | MiniMax M2 Thinking Level · MediumTools | 77.2 | 2025-10-27 | 2300 |
| 19 | Gemma 4 31B Thinking Level · MediumTools | 76.9 | 2026-04-02 | 31 |
| 20 | GLM-4.6 Thinking Level · MediumTools | 75.9 | 2025-09-30 | 3550 |
| 21 | Qwen3 Max (Preview) NormalTools | 74 | 2025-09-05 | 未知 |
| 22 | Claude Opus 4 Thinking Level · MediumTools | 72.5 | 2025-05-23 | 未知 |
| 23 | Qwen3 Max (Preview) Thinking Level · MediumTools | 72 | 2025-09-05 | 未知 |
| 24 | Claude Sonnet 4.5 NormalTools | 71 | 2025-09-30 | 未知 |
| 25 | Gemma 4 26B A4B Thinking Level · MediumTools | 68.2 | 2026-04 | 25.2 |
| 26 | DeepSeek V3.2-Exp Thinking Level · MediumTools | 66.7 | 2025-09-29 | 6710 |
| 27 | Kimi K2 NormalTools | 64.3 | 2025-07-11 | 10000 |
| 28 | Kimi K2 Thinking Level · MediumTools | 64.3 | 2025-07-11 | 10000 |
| 29 | Claude Sonnet 3.7 Thinking Level · MediumTools | 61.8 | 2025-02-25 | 未知 |
| 30 | OpenAI o4 - mini Thinking Level · MediumTools | 56.9 | 2025-04-16 | 未知 |
滚动或悬停加载剩余 10 条