BioSecBench-Function V1
Science
BioSecBench-Function V1: 111 deterministic biological-function evaluation tasks on experimental data. Published pass rates distinguish model and agent harness.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5 | 50.3 |
| GPT-6 Astra | 48.6 |
| Opus 4.8 | 44.6 |
| Grok 4.6 | 44.1 |
| Gemini 3.5 Flash | 42.2 |
| DeepSeek V4.1 Flash | 38.7 |
| GPT-5.6 Sol | 38.3 |
| Sonnet 5 | 37.2 |
| GPT-5.5 | 36.5 |
| Grok 4.5 | 35.6 |
| GPT-5.6 Terra | 33.5 |
| GPT-5.6 Luna | 26.9 |