Expired50% confidenceEventExact time
OpenAI no longer recommends using SWE-Bench Pro as a benchmark for evaluating code generation capabilities.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/10/2026
First Seen
This claim has expired. It is kept for the record and no longer describes the current state.
Valid until: 10/8/2026(expired)
Sources
OpenAI Abandons SWE-Bench Pro: The Trust Crisis in AI Coding Benchmarks and What Comes Next
hackernewshackernews7/8/2026
Related Claims
UnverifiedOrchestration frameworks such as SpecKit, BMAD, and OpenSpec cannot provide a persistent end-to-end environment for running, validating, and managing AI work products at scale.63% similarUnverified在AIME 2024和CodeForces等权威基准测试中，DeepSeek R1的表现与OpenAI O1正式版不相上下62% similarUnverifiedOpenAI introduced Troubleshooting Bench, a benchmark containing real error problems from biological experiment protocols.61% similarUnverifiedOpenAI's and Anthropic's Tool Use protocol formats are not compatible with each other, and best practices for system prompts also differ.59% similarVerifiedOpenAI has adopted an open-weight model rather than open source in the strict sense, without disclosing training data or training code.59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/465769API
curl https://kongchang.com/api/v1/knowledge/claims/465769MCP
get_claim(id=465769)