OpenAI introduces GeneBench-Pro, a research-level benchmark for agentic computational biology
On June 30, 2026, OpenAI announced GeneBench-Pro: a research-level benchmark meant to measure how well AI agents navigate messy biological data and make the judgment calls real computational biology depends on. OpenAI says GeneBench-Pro contains 129 questions and is open-sourcing 10 representative case studies as a public package on Hugging Face under the MIT License.
OpenAI announced GeneBench-Pro on June 30, 2026 as a research-level benchmark aimed at measuring a harder kind of progress than knowledge recall: whether AI agents can navigate messy biological data, choose an analysis path, and make consequential judgment calls that real computational research depends on (OpenAI on X, OpenAI).
The announcement matters if you build or deploy agentic systems for scientific workflows, because it tries to operationalize evaluation beyond one-shot Q&A: each task is framed as a self-contained analysis in an isolated workspace, with data files and a standard bioinformatics stack, and the goal is to test whether an agent can iterate, diagnose, and decide -- not just run a predefined script (OpenAI).
Key Takeaways
- What: GeneBench-Pro is a research-level benchmark for judgment-heavy computational biology agent work (OpenAI).
- What it tries to measure: Handling ambiguity, revising assumptions, choosing analysis paths, and knowing when results are decision-ready (OpenAI).
- Scale: OpenAI says the benchmark contains 129 questions (OpenAI).
- Public access: OpenAI says it is open-sourcing 10 representative questions on Hugging Face, and the dataset page lists 10 problems and states the package is under the MIT License (OpenAI, Hugging Face dataset page).
What OpenAI Says GeneBench-Pro Is Measuring
OpenAI describes GeneBench-Pro as a benchmark for whether models can do the kind of iterative, judgment-heavy analysis common in computational biology: dealing with ambiguous or messy data, deciding what the data can support, and adjusting course when early diagnostics change the right approach (OpenAI).
OpenAI defines this kind of capability in terms of "research taste" -- the sequence of judgment calls that shape an analysis, including which questions the data can support and when an initial plan needs to be revised (OpenAI).
What Is Public Today
OpenAI says it is fully open-sourcing 10 representative GeneBench-Pro questions on Hugging Face (OpenAI).
The public package on Hugging Face lists 10 problem directories under problems/, and includes files like problems.csv, manifest.json, and a reference_grader.py reference grader, and it explicitly states the package is available under the MIT License (Hugging Face dataset page).
How to Think About It (Practically)
If you're building scientific agents, GeneBench-Pro is a signal that evaluation may be shifting toward end-to-end analysis tasks with tool usage and iterative investigation -- not just biology trivia or isolated subtasks (OpenAI).
If you're comparing models or agent configurations internally, the public 10-problem package is a starting point you can integrate into your own harnesses, since it includes reference grading code and a standardized answer schema per problem (Hugging Face dataset page).
Read next
Related coverage- OpenAI
OpenAI Previews GPT-5.6: Sol, Terra, and Luna in Limited Preview
OpenAI previewed the GPT-5.6 family in June 2026 with trusted partners: Sol, a frontier flagship it called a step function better than GPT-5.5; Terra, competitive with GPT-5.5 at 2x lower cost; and Luna, its most cost-efficient model. All three became generally available on July 9, 2026, and GPT-6 has since superseded them.
- OpenAI
OpenAI Introduces LifeVal, a Benchmark for Everyday AI Tasks
OpenAI introduced LifeVal on November 2, 2026, an evaluation of how well AI models help with everyday personal-assistant tasks. Its task areas were shaped by a survey of 3,995 respondents, and OpenAI reports unweighted average-reward scores of 49.3% for GPT-5.2 and 80.6% for GPT-6 Astra.
- OpenAI
OpenAI Reports Stronger Coding Results in COBOL, Fortran, and Visual Basic
OpenAI says its latest models improved on two software-engineering evaluations for low-resource languages. On MultiLangBench, GPT-6 Astra averaged 87.4% at maximum reasoning effort across 20 programming languages; on a COBOL migration test, GPT-6.1 Sol scored 86.6% at low reasoning effort. These are OpenAI-designed evaluations, not a guarantee of production migration quality.
- OpenAI
GPT-6.1 Sol: Astra-Level Coding at a Fifth of the Price, With Cost at Every Effort Level
OpenAI released GPT-6.1 Sol on September 29, 2026, at $2 input and $10 output per million tokens, a fifth of GPT-6 Astra's price, with cached input at $0.10. On OpenAI's charts it edges past Astra's best DeepSWE score at high effort and trails Astra by 2.1 points on OSWorld offline. The API model ID is gpt-6.1-sol.
Frequently Asked Questions
What is GeneBench-Pro?
GeneBench-Pro is a research-level benchmark OpenAI introduced to test how well AI agents can do judgment-heavy computational biology work -- for example, navigating messy biological data, choosing an analysis path, revising assumptions, and deciding when results are decision-ready. OpenAI positions it as an expansion of GeneBench toward harder, more realistic tasks across genomics, quantitative biology, and translational medicine.
How big is the benchmark?
OpenAI says GeneBench-Pro has 129 questions, spanning a range of computational biology settings and methods.
Is anything public/open today?
Yes. OpenAI says it is open-sourcing 10 representative GeneBench-Pro questions on Hugging Face. The public package lists 10 problem directories under `problems/` and states it is available under the MIT License.