Expired50% confidenceFactExact time
A SWE-bench score of 63.2 means the model can correctly solve approximately 63.2% of real software engineering problems.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
This claim has expired. It is kept for the record and no longer describes the current state.
Valid until: 9/30/2026(expired)
Sources
Cursor Composer 2.5: The Secret Behind an Open-Source Model's Reinforcement Training to a Top-3 Coding Benchmark Ranking
bilibili鲲鹏Talk5/20/2026
Related Claims
UnverifiedMuse Glimmer 30B scored 51.2 on the SWE benchmark, meaning it can solve approximately half of real-world software engineering problems.60% similarVerifiedDevin resolved 13.86% of real GitHub Issues in the SWE-bench test, surpassing the performance of Claude 2.58% similarUnverifiedOn the SWE-bench software engineering benchmark, the Claude 3.5 series achieved a significantly higher issue resolution rate than competitors of the same period.57% similarUnverified初代 DeepSeek Math 7B 在 MATH 竞赛基准测试中以70亿参数达到 51.7% 的正确率57% similarUnverifiedo3-mini-high在SWE-Bench上修复了39%的错误，允许调用内部工具后成功率飙升到61%56% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/43943API
curl https://kongchang.com/api/v1/knowledge/claims/43943MCP
get_claim(id=43943)