GPT-5编程测评大反转！表面不及格，实际63.1%的任务没交卷，全算上成绩比Claude高一倍
Quick Answer
The SWE-BENCH PRO benchmark reveals a surprising performance drop for leading AI models, with GPT-5, Claude Opus 4.1, and Gemini 2.5 all failing to exceed a 25% task completion rate.
Quick Take
However, when considering unsubmitted tasks, GPT-5's effective performance is actually 63.1%, significantly outperforming Claude.
Key Points
- GPT-5, Claude Opus 4.1, and Gemini 2.5 all scored below 25% in task completion.
- GPT-5's effective performance, considering unsubmitted tasks, is 63.1%.
- Claude's performance is significantly lower than GPT-5 when accounting for unsubmitted tasks.
- The benchmark highlights a major challenge for leading AI models in software engineering tasks.
- PRO reveals critical insights into AI model performance in programming.
Article Excerpt
From source RSS / original summaryScale AI的新软件工程基准 PRO，出现反转！ 表面上看，“御三家”集体翻车，没一家的解决率超过25%：. GPT-5、Claude Opus 4. 1、Gemini 2. 5分别
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from WebSearch (Tavily)
See more →全球AI芯片峰会，9月上海见！
The 2026 Global AI Chip Summit will take place in Shanghai on September 22-23, focusing on the evolving AI chip landscape, including the shift from training to inference, the rise of diverse chip technologies, and the restructuring of industry competition. Notable speakers include experts from leading universities and companies, discussing advancements in AI chip architecture and applications.