τ²-benchBrowse 284 families
τ²-bench
Can the model plan, use tools, and finish a multi-step task?
Version / measure
- Domain
- Agents
- Ranked models
- 8
- Reported values
- 11
- Reports
- 2 · lab
- Direction
- Higher is better
- Status
- Version not specified
Rankings
Best reported score per model across effort settings
| # | Model | Best setting | Score |
|---|---|---|---|
| 1 | Claude Opus 4.6Anthropic | maxcost n/a2 values · Feb 19 · source | 99.3% |
| 2 | Gemini 3.1 ProGoogle | highcost n/a1 value · Feb 19 · source | 99.3% |
| 3 | GPT-5.2OpenAI | xhighcost n/a2 values · Feb 19 · source | 98.7% |
| 4 | Claude Opus 4.5Anthropic | reportedcost n/a1 value · Feb 12 · source | 98.2% |
| 5 | Gemini 3 ProGoogle | highcost n/a2 values · Feb 19 · source | 98.0% |
| 6 | Claude Sonnet 4.6Anthropic | maxcost n/a1 value · Feb 19 · source | 97.9% |
| 7 | MiniMax M2.5MiniMaxOW | reportedcost n/a1 value · Feb 12 · source | 97.8% |
| 8 | MiniMax M2.1MiniMaxOW | reportedcost n/a1 value · Feb 12 · source | 87.0% |
Effort curve
Score vs. reported cost. Lines connect complete sweeps only.
No cost-linked effort sweep for this version.
Definition and comparison boundarypublic methodology
Can the model plan, use tools, and finish a multi-step task?
Agentic execution across tools, environments, or long-running workflows. Higher is better. The value is the percentage reported in this lab's table.
Harness, tools, prompts, attempt count, and benchmark version can materially change the result. Compare within one reporting context.
Method / source ↗