PinchBench Best Score
Agents · 2026-02-10
PinchBench Best Score reports the best task-performance score observed across a model's evaluated runs. Runtime, cost, and value fields are excluded from this metric.
Top models (higher is better)
| Model | Score |
|---|---|
| Macaron-V1-Venti | 94.0 |
| Qwen3.7-Max | 93.4 |
| Opus 4.8 | 91.8 |
| GPT-5.5 | 89.0 |
| GLM-5.2 | 88.1 |
| MiniMax M3 | 86.1 |
| Gemini 3.1 Pro Preview | 82.9 |