Claude Opus · DeepSWE · v1.1 · 历代记录

按发布时间查看 Claude Opus 的 2 次发布在 DeepSWE · v1.1 的成绩、配置与披露。

仅为直接前代间的唯一、有效精确值计算差值,要求单位、指标与明确记录的配置和条件一致。

Claude Opus 5

· 本次发布有 5 条成绩记录,含 5 条精确值(可含历史记录)

Claude Opus 5 · DeepSWE 成绩与配置
指标与成绩配置与评测条件出处
Score68.8%原文 68.8% · 单位 %
公布 未确认
DeepSWE v1.1 · Max effort
设置
113 long-horizon software-engineering tasks; Figure 8.3.A effort/cost sweep. The figure does not state effort-specific trial counts. 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
未记录
协议
DeepSWE v1.1; §8.3 prose reports the max-effort summary score; Figure 8.3.A max point is 68.8%.
条件
68.8%, average over five trials. Keep distinct effort-sweep points as separate results.
在排名中查看此成绩
Score57.7%原文 57.7% · 单位 %
公布 未确认
DeepSWE v1.1 · Low effort
设置
113 long-horizon software-engineering tasks; Figure 8.3.A effort/cost sweep. The figure does not state effort-specific trial counts. 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
未记录
协议
DeepSWE v1.1; Figure 8.3.A effort/cost sweep.
条件
113 long-horizon coding tasks; explicit figure-labeled effort; figure-specific run count not stated. Keep the chart series distinct from the prose/table max value, which is also 68.8%.
在排名中查看此成绩
Score66.9%原文 66.9% · 单位 %
公布 未确认
DeepSWE v1.1 · Medium effort
设置
113 long-horizon software-engineering tasks; Figure 8.3.A effort/cost sweep. The figure does not state effort-specific trial counts. Explicit adaptive-thinking effort: medium.
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
medium
框架
未记录
协议
DeepSWE v1.1; Figure 8.3.A effort/cost sweep.
条件
113 long-horizon coding tasks; explicit figure-labeled effort; figure-specific run count not stated. Keep the chart series distinct from the prose/table max value, which is also 68.8%.
在排名中查看此成绩
Score68%原文 68% · 单位 %
公布 未确认
DeepSWE v1.1 · High effort
设置
113 long-horizon software-engineering tasks; Figure 8.3.A effort/cost sweep. The figure does not state effort-specific trial counts. 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
未记录
协议
DeepSWE v1.1; Figure 8.3.A effort/cost sweep.
条件
113 long-horizon coding tasks; explicit figure-labeled effort; figure-specific run count not stated. Keep the chart series distinct from the prose/table max value, which is also 68.8%.
在排名中查看此成绩
Score69.7%原文 69.7% · 单位 %
公布 未确认
DeepSWE v1.1 · Xhigh effort
设置
113 long-horizon software-engineering tasks; Figure 8.3.A effort/cost sweep. The figure does not state effort-specific trial counts. 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
未记录
协议
DeepSWE v1.1; Figure 8.3.A effort/cost sweep.
条件
113 long-horizon coding tasks; explicit figure-labeled effort; figure-specific run count not stated. Keep the chart series distinct from the prose/table max value, which is also 68.8%.
在排名中查看此成绩

Claude Opus 5.5

· 本次发布有 1 条成绩记录,含 1 条精确值(可含历史记录)

与直接前代的记录缺少满足差值计算的同条件依据。

Claude Opus 5.5 · DeepSWE 成绩与配置
指标与成绩配置与评测条件出处
Score74.2%原文 74.2% · 单位 %
公布 未确认
System Card max effort (per-result benchmark harnesses)
设置
System Card §8.1 default: adaptive thinking/max effort and default sampling settings unless a section says otherwise. Benchmark tools, evaluator, safeguards, fallback, and repetitions remain per-result conditions; this config represents the stated model effort, not harness equivalence.
未知项
Exact run time is not disclosed.、Per-benchmark tools, safeguards, and repetition counts are specified only where the source states them.
工具
工具条件未记录
推理预算
max
框架
未记录
协议
DeepSWE v1.1; average over five trials; 113 long-horizon tasks.
条件
Tasks were written from scratch to reduce contamination; run date is unknown. Source-wide conditions: DeepSWE v1.1; 113 long-horizon tasks authored from scratch.
在排名中查看此成绩