OpenAI’s BioBasicsBench
Partly confirmed
The takeaway
On 6 Nov 2026 OpenAI announced BioBasicsBench, built with the enzyme-design company Ligo from 100 tasks drawn from real experiments.
Status
- Claim
Partly confirmed
- Our reporting
- Medium confidence
- Importance
- 2 of 5
- Last verified
- 7 November 2026
Your AI and this story
- GPT-6 Astra190 days after its cutoff
- Claude Opus 5.5129 days after its cutoff
- Gemini 4 Argon220 days after its cutoff
- Grok 4.7159 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 129 days before it.
Key facts
- 100 tasks from real experiments; partner Ligo (enzyme design)
- Stated aims: show where models need to improve on everyday lab analysis; track what gains mean for ‘scientific progress, biosafety, and biosecurity’
- The blog post could not be read in this run (HTTP 403), so model scores are not recorded here
What happened
OpenAI published the benchmark on its blog and on X. This entry is written from the X post only.
Why it matters
It is a lab-built measurement of agents on routine biology analysis, the kind of capability that lab safety frameworks track.
Sources
2 sources from 2 sites. Numbers match the chips in the text.
2 sources: 2 primary
Primary
- OpenAI: Measuring how AI agents handle everyday wet-lab scienceopenai.com, official
- OpenAI on X (~658k views)x.com, official
Changes
- Filed (lab X list)