← All benchmarks

OpenAI Expert-SWE - internal release score

Code

Internal expert software-engineering evaluation. Undisclosed task roster and scoring make the release result diagnostic, not a SWE-bench Pro measurement.

Top models (higher is better)