Claude Opus 5 · Frontier-Bench(v0.1) · 附件证据(ev-op5-card-frontierbench-miniswe)

Claude Opus 5 的 Frontier-Bench · v0.1 官方出处。见《Claude Opus 5 System Card》System Card PDF pp.155-157, §8.5, FrontierBench v0.1 results。证据 ev-op5-card-frontierbench-miniswe。

原文与定位

Claude Opus 5 · Frontier-Bench(v0.1),FrontierBench v0.1 · mini-SWE-agent/GKE · Low effort配置,Score 25%(具体成绩公布时间未确认)。见《Claude Opus 5 System Card》System Card PDF pp.155-157, §8.5, FrontierBench v0.1 results。 查看成绩条件

Claude Opus 5 · Frontier-Bench(v0.1),FrontierBench v0.1 · mini-SWE-agent/GKE · High effort配置,Score 39%(具体成绩公布时间未确认)。见《Claude Opus 5 System Card》System Card PDF pp.155-157, §8.5, FrontierBench v0.1 results。 查看成绩条件

Claude Opus 5 · Frontier-Bench(v0.1),FrontierBench v0.1 · mini-SWE-agent/GKE · Xhigh effort配置,Score 44.4%(具体成绩公布时间未确认)。见《Claude Opus 5 System Card》System Card PDF pp.155-157, §8.5, FrontierBench v0.1 results。 查看成绩条件

Claude Opus 5 · Frontier-Bench(v0.1),FrontierBench v0.1 · mini-SWE-agent/GKE · Max effort配置,Score 43%(具体成绩公布时间未确认)。见《Claude Opus 5 System Card》System Card PDF pp.155-157, §8.5, FrontierBench v0.1 results。 查看成绩条件

Anthropic mini-SWE-agent on GKE: xhigh 44.4% mean reward; max 43%; high 39%; low 25%. Each figure is averaged over five attempts on each of 74 tasks. Opus 4.8 fallback is reported for classifier refusals.
官方材料
Claude Opus 5 System Card · PDF 第 156 页
定位
附件 · System Card PDF pp.155-157, §8.5, FrontierBench v0.1 results
材料版本
Current saved PDF/TXT: 本地留存副本 and current-opus-5-system-card.txt; cover dated 2026-07-24; p.2 changelog records an update on 2026-08-19; later revision date, if any, unknown.
版本公布
未确认
披露时间
未确认 · 首发/后补未确认
采集时间
历史依据
暂无可靠历史版本依据
引用文本

Claude Opus 5 · Frontier-Bench(v0.1),FrontierBench v0.1 · mini-SWE-agent/GKE · Low effort配置,Score 25%(具体成绩公布时间未确认);Claude Opus 5 · Frontier-Bench(v0.1),FrontierBench v0.1 · mini-SWE-agent/GKE · High effort配置,Score 39%(具体成绩公布时间未确认);Claude Opus 5 · Frontier-Bench(v0.1),FrontierBench v0.1 · mini-SWE-agent/GKE · Xhigh effort配置,Score 44.4%(具体成绩公布时间未确认);Claude Opus 5 · Frontier-Bench(v0.1),FrontierBench v0.1 · mini-SWE-agent/GKE · Max effort配置,Score 43%(具体成绩公布时间未确认)。官方出处 Claude Opus 5 System Card(附件,Current saved PDF/TXT: 本地留存副本 and current-opus-5-system-card.txt; cover dated 2026-07-24; p.2 changelog records an update on 2026-08-19; later revision date, if any, unknown.)System Card PDF pp.155-157, §8.5, FrontierBench v0.1 results,https://www.anthropic.com/claude-opus-5-system-card#page=156。Benchlog https://benchlog.lekshome.cc/evidence/claude-opus-5/ev-op5-card-frontierbench-miniswe/

关联成绩与条件

Claude Opus 5 · Frontier-Bench 成绩与配置
指标与成绩配置与评测条件证据关系
Score25%原文 25% · 单位 %
公布 未确认
FrontierBench v0.1 · mini-SWE-agent/GKE · Low effort
设置
Anthropic-run mini-SWE-agent harness on GKE; mean reward over five attempts for each of 74 unique tasks; production safety classifiers were enabled. Refusals triggered fallback to Opus 4.8 in 4% of total trials. Explicit adaptive-thinking effort: low.
未知项
The card does not state any additional sampling parameters beyond the cited evaluation description.
工具
工具条件未记录
推理预算
low
框架
mini-SWE-agent harness
协议
FrontierBench v0.1; Anthropic mini-SWE-agent harness on GKE; mean reward over 74 tasks, five attempts each.
条件
Internal harness run separate from the Harbor Table 8.1.A row; max was reported as similar to xhigh and within noise. Safety classifiers/fallback were active.

直接支持该成绩

在排名中查看此成绩
Score39%原文 39% · 单位 %
公布 未确认
FrontierBench v0.1 · mini-SWE-agent/GKE · High effort
设置
Anthropic-run mini-SWE-agent harness on GKE; mean reward over five attempts for each of 74 unique tasks; production safety classifiers were enabled. Refusals triggered fallback to Opus 4.8 in 4% of total trials. Explicit adaptive-thinking effort: high.
未知项
The card does not state any additional sampling parameters beyond the cited evaluation description.
工具
工具条件未记录
推理预算
high
框架
mini-SWE-agent harness
协议
FrontierBench v0.1; Anthropic mini-SWE-agent harness on GKE; mean reward over 74 tasks, five attempts each.
条件
Internal harness run separate from the Harbor Table 8.1.A row; max was reported as similar to xhigh and within noise. Safety classifiers/fallback were active.

直接支持该成绩

在排名中查看此成绩
Score44.4%原文 44.4% · 单位 %
公布 未确认
FrontierBench v0.1 · mini-SWE-agent/GKE · Xhigh effort
设置
Anthropic-run mini-SWE-agent harness on GKE; mean reward over five attempts for each of 74 unique tasks; production safety classifiers were enabled. Refusals triggered fallback to Opus 4.8 in 4% of total trials. Explicit adaptive-thinking effort: xhigh.
未知项
The card does not state any additional sampling parameters beyond the cited evaluation description.
工具
工具条件未记录
推理预算
xhigh
框架
mini-SWE-agent harness
协议
FrontierBench v0.1; Anthropic mini-SWE-agent harness on GKE; mean reward over 74 tasks, five attempts each.
条件
Internal harness run separate from the Harbor Table 8.1.A row; max was reported as similar to xhigh and within noise. Safety classifiers/fallback were active.

直接支持该成绩

在排名中查看此成绩
Score43%原文 43% · 单位 %
公布 未确认
FrontierBench v0.1 · mini-SWE-agent/GKE · Max effort
设置
Anthropic-run mini-SWE-agent harness on GKE; mean reward over five attempts for each of 74 unique tasks; production safety classifiers were enabled. Refusals triggered fallback to Opus 4.8 in 4% of total trials. Explicit adaptive-thinking effort: max.
未知项
The card does not state any additional sampling parameters beyond the cited evaluation description.
工具
工具条件未记录
推理预算
max
框架
mini-SWE-agent harness
协议
FrontierBench v0.1; Anthropic mini-SWE-agent harness on GKE; mean reward over 74 tasks, five attempts each.
条件
Internal harness run separate from the Harbor Table 8.1.A row; max was reported as similar to xhigh and within noise. Safety classifiers/fallback were active.

直接支持该成绩

在排名中查看此成绩