Lead story
Models & availability
Latest
← Feed
Copy✕ Close copy menu
Why we no longer evaluate SWE-bench Verified
OpenAI explains that they no longer evaluate SWE-bench Verified due to contamination and flawed tests, and they recommend SWE-bench Pro.
From the source
SWE-bench Verified is increasingly contaminated and mismeasures frontier coding progress.
openai.com
Visit source ↗openai.com
- Announced
- 23 Feb 2026 11:00 UTC
- Detected
- 03 Aug 2026 21:11 UTC (+161d after)