CLAW-Eval: leaderboard
CLAW-Eval — agentic tool-use evaluation across realistic multi-step workflows; overall task score.
Metric: Score (%). Source: claw-eval.github.io. Saturation forecast: Around January 2027. 26 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.6 | 81.4 | #90 |
| 2 | MiMo-V2.5-Pro | 80.6 | #158 |
| 3 | Step 3.7 Flash | 80.5 | #236 |
| 4 | Claude Opus 4.6 | 80.4 | #63 |
| 5 | GPT-5.4 | 78.4 | #79 |
| 6 | MiMo-V2.5 | 78.4 | #196 |
| 7 | Kimi K2.6 | 77.9 | #104 |
| 8 | Gemini 3.1 Pro (Preview) | 77.3 | #58 |
| 9 | GLM-5.1 | 77.3 | #97 |
| 10 | DeepSeek V4 Pro | 77.2 | #102 |
| 11 | MiMo-V2-Pro | 77 | #168 |
| 12 | Muse Spark | 76.3 | #64 |
| 13 | Qwen 3.6 Plus | 75 | #120 |
| 14 | Kimi K2.5 | 74.9 | #144 |
| 15 | GLM-5-Turbo | 74.4 | #119 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=claw-eval · How It Works · Data refreshed daily, snapshot 2026-10-20.