Toolathlon-VerifiedBrowse 284 families
Toolathlon-Verified
Can the model plan, use tools, and finish a multi-step task?
Version / measure
- Domain
- Agents
- Ranked models
- 12
- Reported values
- 16
- Reports
- 3 · lab
- Direction
- Higher is better
- Status
- Version not specified
Rankings
Best reported score per model across effort settings
| # | Model | Best setting | Score |
|---|---|---|---|
| 1 | Claude Fable 5Anthropic | maxcost n/a1 value · Jul 17 · source | 77.9% |
| 2 | Claude Opus 4.8Anthropic | reportedcost n/a3 values · Jul 31 · source | 76.2% |
| 3 | Muse Spark 1.1Meta | xhighcost n/a1 value · Jul 9 · source | 75.6% |
| 4 | GPT-5.6 SolOpenAI | maxcost n/a1 value · Jul 17 · source | 74.9% |
| 5 | GPT-5.5OpenAI | xhighcost n/a2 values · Jul 17 · source | 73.5% |
| 6 | Kimi K3Moonshot AIOW | maxcost n/a1 value · Jul 17 · source | 73.2% |
| 7 | DeepSeek V4 Flash 0731DeepSeekOW | maxcost n/a1 value · Jul 31 · source | 70.3% |
| 8 | Gemini 3.1 ProGoogle | highcost n/a1 value · Jul 9 · source | 61.1% |
| 9 | GLM-5.2Z.aiOW | reportedcost n/a2 values · Jul 31 · source | 59.9% |
| 10 | DeepSeek V4 ProDeepSeekOW | reportedcost n/a1 value · Jul 31 · source | 55.9% |
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.
108-task current release using the official Toolathlon harness.
Method / source ↗