Decision-Bench v1.1 Solved
Professional Work · 2026-10-01
Share of the same 210 Decision-Bench v1.1 tasks on which a run scores at least 95 of 100. A thresholded view of the mean task score, kept as a diagnostic so it never enters capability.
Top models (higher is better)
| Model | Score |
|---|---|
| Gemini 4 Argon | 35.2 |
| Claude Opus 5.5 | 34.3 |
| Claude Sonnet 5.5 | 28.1 |
| GPT-6 Astra | 27.1 |
| GPT-6.1 Sol | 24.3 |
| DeepSeek V4.1 Flash | 17.6 |
| GPT-6 Sol | 17.6 |
| Muse Wonder | 17.6 |
| Gemini 3.8 Flash | 16.2 |
| Claude Haiku 5.5 | 13.8 |
| Kimi K3 | 13.8 |
| GLM-5.3 Flash | 12.4 |
| Muse Spark 1.3 | 11.9 |
| Sonnet 5 | 9.1 |
| GPT-6 Luna | 7.1 |
| Haiku 4.5 | 0.9 |