PinchBench - 147 Tasks - OpenClaw 2026.5.4 (325df3e): leaderboard
Metric: Best Official Run Score (%). Source: pinchbench.com. Saturation forecast: Not forecast. 48 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3.7 Max | 93.44 |
| 2 | Grok Build 0.1 | 92.07 |
| 3 | Claude Opus 4.8 | 91.76 |
| 4 | Claude Opus 4.7 | 91.58 |
| 5 | MiMo-V2.5 | 91.42 |
| 6 | Claude Haiku 4.5 | 90.4 |
| 7 | Seed 2.0 Lite | 89.71 |
| 8 | MiMo-V2.5-Pro | 89.52 |
| 9 | DeepSeek V4 Flash | 89.52 |
| 10 | Qwen 3.6 Flash | 89.14 |
| 11 | GPT-5.5 | 88.96 |
| 12 | Claude Opus 4.6 | 88.89 |
| 13 | GPT-5.4 | 88.39 |
| 14 | GLM-5-Turbo | 88.2 |
| 15 | GLM-5.2 | 87.79 |
Interactive version: theaggregate.ai/benchmark?slug=pinchbench-147-tasks-openclaw-2026-5-4-325df3e · How It Works · Data refreshed daily, snapshot 2026-10-09.