TxBench Antibody Discovery V2
Science
TxBench-Antibody-Discovery V2 clean-v1 reruns on 137 experimental-data evaluations. Macro-mean pass rate; the aggregate site's weight of 100 is not the task count.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-6 Astra | 48.9 |
| Grok 4.6 | 44.5 |
| Gemini 3.7 Flash | 43.5 |
| Gemini 3.8 Flash | 43.3 |
| Grok 4.7 | 42.8 |
| Claude Opus 5 | 41.9 |
| GPT-6 Sol | 37.0 |
| Opus 4.7 | 32.9 |
| Opus 4.8 | 32.4 |
| GPT-5.6 Sol | 31.4 |
| Sonnet 5 | 30.4 |
| GPT-5.5 | 29.9 |
| GPT-6 Luna | 29.9 |
| Sonnet 4.6 | 23.4 |
| GPT-5.6 Luna | 19.2 |
| Claude Opus 5.5 | 6.8 |