theaicatalog
OpenAI ยท Research

OpenAI introduces benchmark for AI agents on routine wet-lab work

OpenAI has introduced BioBasicsBench, a benchmark that tests how well AI agents interpret raw experimental data across 100 routine wet-lab tasks. The benchmark was developed with Ligo, an enzyme design biotech company, and covers areas such as protein methods and molecular cloning. At the xhigh reasoning setting, GPT-6 Astra earned an average of 66.7% of the available credit, compared with 43.8% for GPT-5.6 Sol. OpenAI says the results show progress on routine lab analysis but room to improve in interpreting images and checking controls.