OpenAI introduces benchmark for routine lab work
OpenAI introduced BioBasicsBench, a benchmark for evaluating agents on routine laboratory analysis.
- BioBasicsBench contains 100 tasks based on raw experimental laboratory data.
- OpenAI reports 66.7% of available credit for GPT-6 Astra and 43.8% for GPT-5.6 Sol at xhigh.
- The evaluation still finds limits in image reading, controls and experimental context.
Context and analysis by Noosigma AI
Editorial summary and context based on the cited source. Claims and results attributed to the organization are reports from that source; they do not constitute independent validation by Noosigma AI.
Its 100 tasks test whether agents can interpret raw experimental evidence, identify errors and recognize when the data do not support a conclusion. Examples include gel images, microscopy, instrument readouts and protocol records.
Why it matters
In the published evaluation, GPT-6 Astra at the xhigh reasoning setting earned an average of 66.7% of available credit, compared with 43.8% for GPT-5.6 Sol. OpenAI developed the benchmark with Ligo, a biotechnology company focused on enzyme design.
These results come from OpenAI’s own evaluation. The article still identifies weaknesses in reading images, checking controls and applying experimental context, so the benchmark does not show that agents can replace laboratory work. Source: OpenAI, November 6, 2026.