BioSecBench-Surveillance V2
Science
BioSecBench-Surveillance V2 clean-v1 reruns on 102 genomic surveillance evaluations. Macro-mean pass rate on updated evaluation sets, distinct from V1.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5.5 | 58.5 |
| Claude Opus 5 | 51.5 |
| GPT-6 Astra | 48.5 |
| Grok 4.6 | 48.0 |
| Grok 4.7 | 48.0 |
| GPT-6 Sol | 47.7 |
| GPT-5.6 Sol | 47.3 |
| GPT-5.5 | 47.1 |
| Opus 4.8 | 46.9 |
| Opus 4.7 | 45.8 |
| Gemini 3.8 Flash | 44.1 |
| GPT-5.6 Luna | 43.2 |
| Gemini 3.7 Flash | 41.5 |
| GPT-6 Luna | 41.5 |
| Sonnet 5 | 39.0 |
| Sonnet 4.6 | 34.5 |