micro1 PharmaBench Mean Score
Health & Medicine
micro1 PharmaBench's mean score over the same 120 tasks: each run earns credit for the required criteria it meets less the weight of the penalties it triggers, averaged over a task's three runs and then over tasks. It shows how close runs that do not pass come to a complete answer.
Top models (higher is better)
| Model | Score |
|---|---|
| Kimi K3.1 Pro | 81.8 |
| Muse Wonder | 80.9 |
| GPT-6 Astra | 78.6 |
| Grok 4.6 | 77.1 |
| Muse Spark 1.3 | 77.1 |
| Claude Sonnet 5.5 | 75.6 |
| Claude Fable 5.1 | 75.5 |
| Claude Opus 5.5 | 75.1 |
| Gemini 4 Argon | 73.6 |
| Claude Opus 5 | 73.2 |
| Muse Spark 1.1 | 70.5 |
| Gemini 3.8 Flash | 66.6 |