LifeSciBench - release-reported version unspecified
Science
LifeSciBench release score without an identified Gold version or task roster. Diagnostic until equivalence with the named Gold v1 cohort is established.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.6 Sol | 59.9 |
| GPT-5.6 Terra | 56.0 |
| GPT-5.6 Luna | 51.2 |