GeneBench · 2026-04-23 report; 103 evaluations across 10 domains · 测试定义与成绩

GeneBench · 2026-04-23 report; 103 evaluations across 10 domains 收录 1 次模型发布的评测披露或成绩。

测试版本与指标

  • Mean pass rate · % · 越高越好

各系列最新模型 · 公布成绩

常规展示范围。排名按指标与评测协议分组,已记录的工具、预算及框架分别展示。条件未说明的组保留公布值供查证。

Mean pass rate %

允许工具 / xhigh / 框架未记录 · GeneBench full suite; 103 evaluations across 10 domains; mean unweighted per-problem pass rate over repeated runs; 95% CI in source excerpt

GeneBench · 2026-04-23 report; 103 evaluations across 10 domains · Mean pass rate · 允许工具 / xhigh / 框架未记录 · GeneBench full suite; 103 evaluations across 10 domains; mean unweighted per-problem pass rate over repeated runs; 95% CI in source excerpt
模型与配置成绩(%)记录与证据
GPT-5.5Xhigh25%

允许工具 / 推理预算未记录 / Pro harness · GeneBench full suite; separately reported GPT Pro harness; mean unweighted per-problem pass rate over repeated runs; 95% CI in source excerpt

GeneBench · 2026-04-23 report; 103 evaluations across 10 domains · Mean pass rate · 允许工具 / 推理预算未记录 / Pro harness · GeneBench full suite; separately reported GPT Pro harness; mean unweighted per-problem pass rate over repeated runs; 95% CI in source excerpt
模型与配置成绩(%)记录与证据
GPT-5.5Pro33.2%

允许工具 / none / 框架未记录 · GeneBench full suite; 103 evaluations across 10 domains; mean unweighted per-problem pass rate over repeated runs; 95% CI in source excerpt

GeneBench · 2026-04-23 report; 103 evaluations across 10 domains · Mean pass rate · 允许工具 / none / 框架未记录 · GeneBench full suite; 103 evaluations across 10 domains; mean unweighted per-problem pass rate over repeated runs; 95% CI in source excerpt
模型与配置成绩(%)记录与证据
GPT-5.5None1.9%

允许工具 / low / 框架未记录 · GeneBench full suite; 103 evaluations across 10 domains; mean unweighted per-problem pass rate over repeated runs; 95% CI in source excerpt

GeneBench · 2026-04-23 report; 103 evaluations across 10 domains · Mean pass rate · 允许工具 / low / 框架未记录 · GeneBench full suite; 103 evaluations across 10 domains; mean unweighted per-problem pass rate over repeated runs; 95% CI in source excerpt
模型与配置成绩(%)记录与证据
GPT-5.5Low3.2%

允许工具 / medium / 框架未记录 · GeneBench full suite; 103 evaluations across 10 domains; mean unweighted per-problem pass rate over repeated runs; 95% CI in source excerpt

GeneBench · 2026-04-23 report; 103 evaluations across 10 domains · Mean pass rate · 允许工具 / medium / 框架未记录 · GeneBench full suite; 103 evaluations across 10 domains; mean unweighted per-problem pass rate over repeated runs; 95% CI in source excerpt
模型与配置成绩(%)记录与证据
GPT-5.5Medium9.2%

允许工具 / high / 框架未记录 · GeneBench full suite; 103 evaluations across 10 domains; mean unweighted per-problem pass rate over repeated runs; 95% CI in source excerpt

GeneBench · 2026-04-23 report; 103 evaluations across 10 domains · Mean pass rate · 允许工具 / high / 框架未记录 · GeneBench full suite; 103 evaluations across 10 domains; mean unweighted per-problem pass rate over repeated runs; 95% CI in source excerpt
模型与配置成绩(%)记录与证据
GPT-5.5High22.2%

允许工具 / low / 框架未记录 · 条件未说明

GeneBench · 2026-04-23 report; 103 evaluations across 10 domains · Mean pass rate · 允许工具 / low / 框架未记录 · 条件未说明
模型与配置成绩(%)记录与证据
GPT-5.5Low3.1%

历史记录 · 1 次发布

  • GPT-5.5本次发布有 7 条成绩记录,含 7 条精确值(可含历史记录)