OpenAI introduces BioBasicsBench for AI lab analysis
Original title: Measuring how AI agents handle everyday wet-lab science
OpenAI released BioBasicsBench benchmark to test AI agents on 100 routine wet-lab science tasks.
OpenAI introduced BioBasicsBench, a benchmark developed with Ligo biotech company to evaluate AI agents on routine wet-lab tasks. The 100 tasks test interpretation of raw experimental data including gel images, microscopy, and protocols. GPT-6 Astra achieved 66.7% accuracy at high reasoning setting compared to 43.8% for GPT-5.6 Sol. The benchmark covers protein methods, imaging, assays, nucleic acids, and molecular cloning.
- BioBasicsBench includes 100 tasks across five domains
- GPT-6 Astra scores 66.7% versus 43.8% for GPT-5.6 Sol
- Tasks use raw experimental evidence from real labs
- Agents must detect errors and recommend actions
- AI models
- GPT-6 AstraGPT-5.6 Sol
From the original article
11 min read at openai.comOpenAI introduces BioBasicsBench, a 100-task benchmark measuring how well AI agents interpret raw experimental data and handle everyday wet-lab science.
openai.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.