RealClawBench - Pass@3: leaderboard
Metric: Pass@3 (%): share of the 281 tasks solved in at least one of three runs, on 281 released tasks reconstructed from real OpenClaw developer-agent sessions, run in the shared OpenClaw agent runtime (file read, write, edit, search and shell tools, 600-second timeout, temperature 1.0) and scored by case-specific deterministic Python verifiers; mean of three independent runs; higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 14 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | DeepSeek V4 Pro | 76.2 |
| 2 | GPT-5.5 | 75.1 |
| 3 | Claude Opus 4.7 | 75.1 |
| 4 | MiMo-V2.5-Pro | 74.7 |
| 5 | Kimi K2.6 | 73 |
| 6 | DeepSeek V4 Flash | 71.2 |
| 7 | GLM-5.1 | 70.8 |
| 8 | Claude Sonnet 4.6 | 70.5 |
| 9 | Gemini 3.1 Pro (Preview) | 69 |
| 10 | MiniMax-M2.7 | 68.3 |
| 11 | Qwen 3.6 Plus | 67.6 |
| 12 | Claude Opus 4.6 | 67.3 |
| 13 | GPT-OSS-120B | 57.7 |
| 14 | Gemma 4 31B | 51.6 |
Interactive version: theaggregate.ai/benchmark?slug=realclawbench-pass-3 · How It Works · Data refreshed daily, snapshot 2026-09-29.