BioBasicsBench measures agents on routine wet-lab tasks
On 6 November 2026 OpenAI published BioBasicsBench with Ligo. One hundred tasks test whether an agent can read raw lab evidence, spot an error, and stop when the evidence does not support a conclusion.
OpenAI published BioBasicsBench on 6 November 2026. It is a benchmark, built with Ligo, a biotech company focused on enzyme design. It is not a new model. The page says the 100 tasks test whether an agent can interpret raw experimental data, identify errors, and recognize when the evidence cannot support a conclusion.
At the xhigh setting, which the page describes as more reasoning per task, GPT-6 Astra earned an average of 66.7% of the available credit. GPT-5.6 Sol earned 43.8%. The page calls that progress on routine lab analysis, with room to improve on images, controls, and experimental context. It does not say a higher score means the model can work in a lab.
The 100 tasks cover five areas: protein methods, imaging and cytometry, quantification and assays, nucleic acid methods, and molecular cloning. Scientists built the tasks from their own experiments and paired each question with expert-written grading criteria. Each task had an internal review by a Ligo scientist who had not written it, an external review by a PhD-level scientist working in a wet lab, and a final internal review.
The page groups what the agent is given. Gels and blots appear in 31 tasks. Protocols, notes, and specs in 27. Tables and instrument exports in 20. Bench, apparatus, and reagent photos in 18. Microscopy in 14. Instrument screens and plots in 13. Sequencing files in 12. Protocol videos in 7. Agar plates and colonies in 7. Reference sheets in 21. Those rows add to 170, so they are not a split of the 100 tasks. The page extract did not say whether one task can sit in more than one row.
What this desk did not open
The page extract cut off before a full results table. This draft does not report token counts, confidence intervals, or scores for models other than the two in the opening paragraph. It also does not describe how any lab method is performed. The post of @OpenAI that pointed at the page was not opened as a page. The facts above are from the OpenAI page.