GPT-5.1 Codex · SWE-bench Verified(Verified subset (500 problems)) 73.7% · 主文证据(ev-gpt51-codex-swe)

GPT-5.1 Codex 的 SWE-bench Verified · Verified subset (500 problems) 官方出处。GPT-5.1 Codex (high) · 73.7%。见《构建更多可能:GPT-5.1-Codex-Max 简介 | OpenAI》Appendix: Model evaluations / SWE-bench Verified (n=500), GPT-5.1-Codex column。证据 ev-gpt51-codex-swe。

原文与定位

GPT-5.1 Codex · SWE-bench Verified(Verified subset (500 problems)),GPT-5.1 Codex (high)配置,Pass rate 73.7%(成绩公布 2025-11-19)。见《构建更多可能:GPT-5.1-Codex-Max 简介 | OpenAI》Appendix: Model evaluations / SWE-bench Verified (n=500), GPT-5.1-Codex column。 查看成绩条件

SWE-bench Verified (n=500): GPT-5.1-Codex (high) 73.7%.
官方材料
构建更多可能:GPT-5.1-Codex-Max 简介 | OpenAI
定位
主文 · Appendix: Model evaluations / SWE-bench Verified (n=500), GPT-5.1-Codex column
材料版本
Official GPT-5.1-Codex-Max release article; first same-day CDX capture 2025-11-19 20:09:39 UTC, status 200 text/html; benchmark comparison table read.
版本公布
披露时间
· 后补披露
采集时间
历史依据
Wayback CDX capture 20251119200939, status 200 text/html; the local archived body for that capture was read. A fresh replay request was rate-limited with HTTP 429.
引用文本

GPT-5.1 Codex · SWE-bench Verified(Verified subset (500 problems)),GPT-5.1 Codex (high)配置,Pass rate 73.7%(成绩公布 2025-11-19)。官方出处 构建更多可能:GPT-5.1-Codex-Max 简介 | OpenAI(主文,Official GPT-5.1-Codex-Max release article; first same-day CDX capture 2025-11-19 20:09:39 UTC, status 200 text/html; benchmark comparison table read.)Appendix: Model evaluations / SWE-bench Verified (n=500), GPT-5.1-Codex column,https://openai.com/zh-Hans-CN/index/gpt-5-1-codex-max/。Benchlog https://benchlog.lekshome.cc/evidence/gpt-5-1-codex/ev-gpt51-codex-swe/

关联成绩与条件

GPT-5.1 Codex · SWE-bench Verified 成绩与配置
指标与成绩配置与评测条件证据关系
Pass rate73.7%原文 73.7% · 单位 %
公布
GPT-5.1 Codex (high)
设置
The later Codex-Max evaluation table explicitly labels this comparison column high.
未知项
Codex-Max article footnote says all evaluation runs used compaction and extra-high reasoning, while the table labels this comparison column high; unresolved discrepancy is preserved.
工具
工具条件未记录
推理预算
未记录
框架
未记录
协议
SWE-bench Verified, n=500.
条件
Table labels high effort; source footnote also says evaluation used compaction and extra-high reasoning, unresolved.

直接支持该成绩

在排名中查看此成绩