WANDR · 500-task wide research benchmark · 测试定义与成绩

WANDR · 500-task wide research benchmark 收录 4 次模型发布的评测披露或成绩。

测试版本与指标

别名 Wide ANd Deep Research

  • Soft F1 · % · 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。

Soft F1 %

允许工具 / low / 框架未记录 · WANDR 500-task variant; low effort; web search/fetch; 980k-token budget; soft F1 averaged over tasks.

WANDR · 500-task wide research benchmark · Soft F1 · 允许工具 / low / 框架未记录 · WANDR 500-task variant; low effort; web search/fetch; 980k-token budget; soft F1 averaged over tasks.
模型与配置成绩(%)记录与证据
Claude Fable 5.2Low71.4%

允许工具 / medium / 框架未记录 · WANDR 500-task variant; medium effort; web search/fetch; 980k-token budget; soft F1 averaged over tasks.

WANDR · 500-task wide research benchmark · Soft F1 · 允许工具 / medium / 框架未记录 · WANDR 500-task variant; medium effort; web search/fetch; 980k-token budget; soft F1 averaged over tasks.
模型与配置成绩(%)记录与证据
Claude Fable 5.2Medium73.6%

允许工具 / high / 框架未记录 · WANDR 500-task variant; high effort; web search/fetch; 980k-token budget; soft F1 averaged over tasks.

WANDR · 500-task wide research benchmark · Soft F1 · 允许工具 / high / 框架未记录 · WANDR 500-task variant; high effort; web search/fetch; 980k-token budget; soft F1 averaged over tasks.
模型与配置成绩(%)记录与证据
Claude Fable 5.2High79.3%

允许工具 / xhigh / 框架未记录 · WANDR 500-task variant; xhigh effort; web search/fetch; 980k-token budget; soft F1 averaged over tasks.

WANDR · 500-task wide research benchmark · Soft F1 · 允许工具 / xhigh / 框架未记录 · WANDR 500-task variant; xhigh effort; web search/fetch; 980k-token budget; soft F1 averaged over tasks.
模型与配置成绩(%)记录与证据
Claude Fable 5.2Xhigh80.6%

允许工具 / max / 框架未记录 · WANDR 500-task variant; max effort; web search/fetch; 980k-token budget; soft F1 averaged over tasks.

WANDR · 500-task wide research benchmark · Soft F1 · 允许工具 / max / 框架未记录 · WANDR 500-task variant; max effort; web search/fetch; 980k-token budget; soft F1 averaged over tasks.
模型与配置成绩(%)记录与证据
Claude Fable 5.2Max82.1%

允许工具 / low / 框架未记录 · WANDR; 500 wide-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; soft F1 averaged across tasks.

WANDR · 500-task wide research benchmark · Soft F1 · 允许工具 / low / 框架未记录 · WANDR; 500 wide-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; soft F1 averaged across tasks.
模型与配置成绩(%)记录与证据
Claude Haiku 5.5Low3.5%

允许工具 / medium / 框架未记录 · WANDR; 500 wide-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; soft F1 averaged across tasks.

WANDR · 500-task wide research benchmark · Soft F1 · 允许工具 / medium / 框架未记录 · WANDR; 500 wide-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; soft F1 averaged across tasks.
模型与配置成绩(%)记录与证据
Claude Haiku 5.5Medium12.9%

允许工具 / high / 框架未记录 · WANDR; 500 wide-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; soft F1 averaged across tasks.

WANDR · 500-task wide research benchmark · Soft F1 · 允许工具 / high / 框架未记录 · WANDR; 500 wide-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; soft F1 averaged across tasks.
模型与配置成绩(%)记录与证据
Claude Haiku 5.5High37.3%

允许工具 / xhigh / 框架未记录 · WANDR; 500 wide-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; soft F1 averaged across tasks.

WANDR · 500-task wide research benchmark · Soft F1 · 允许工具 / xhigh / 框架未记录 · WANDR; 500 wide-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; soft F1 averaged across tasks.
模型与配置成绩(%)记录与证据
Claude Haiku 5.5Xhigh45.9%

允许工具 / max / 框架未记录 · WANDR; 500 wide-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; soft F1 averaged across tasks.

WANDR · 500-task wide research benchmark · Soft F1 · 允许工具 / max / 框架未记录 · WANDR; 500 wide-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; soft F1 averaged across tasks.
模型与配置成绩(%)记录与证据
Claude Haiku 5.5Max49.9%

允许工具 / low / 框架未记录 · WANDR, 500 tasks; soft F1 averaged across tasks.

WANDR · 500-task wide research benchmark · Soft F1 · 允许工具 / low / 框架未记录 · WANDR, 500 tasks; soft F1 averaged across tasks.
模型与配置成绩(%)记录与证据
Claude Sonnet 5.5Low10%

允许工具 / med / 框架未记录 · WANDR, 500 tasks; soft F1 averaged across tasks.

WANDR · 500-task wide research benchmark · Soft F1 · 允许工具 / med / 框架未记录 · WANDR, 500 tasks; soft F1 averaged across tasks.
模型与配置成绩(%)记录与证据
Claude Sonnet 5.5Medium29.9%

允许工具 / high / 框架未记录 · WANDR, 500 tasks; soft F1 averaged across tasks.

WANDR · 500-task wide research benchmark · Soft F1 · 允许工具 / high / 框架未记录 · WANDR, 500 tasks; soft F1 averaged across tasks.
模型与配置成绩(%)记录与证据
Claude Sonnet 5.5High56.3%

允许工具 / xhigh / 框架未记录 · WANDR, 500 tasks; soft F1 averaged across tasks.

WANDR · 500-task wide research benchmark · Soft F1 · 允许工具 / xhigh / 框架未记录 · WANDR, 500 tasks; soft F1 averaged across tasks.
模型与配置成绩(%)记录与证据
Claude Sonnet 5.5Xhigh66.6%

允许工具 / max / 框架未记录 · WANDR, 500 tasks; soft F1 averaged across tasks.

WANDR · 500-task wide research benchmark · Soft F1 · 允许工具 / max / 框架未记录 · WANDR, 500 tasks; soft F1 averaged across tasks.
模型与配置成绩(%)记录与证据
Claude Sonnet 5.5Max70%

允许工具 / low / 框架未记录 · WANDR; low effort; web search/fetch on frozen index, programmatic tools, code execution; 980k-token budget; five-hundred-task mean; Claude Opus 4.6 grader.

WANDR · 500-task wide research benchmark · Soft F1 · 允许工具 / low / 框架未记录 · WANDR; low effort; web search/fetch on frozen index, programmatic tools, code execution; 980k-token budget; five-hundred-task mean; Claude Opus 4.6 grader.
模型与配置成绩(%)记录与证据
Claude Opus 5.5Low31.2%

允许工具 / medium / 框架未记录 · WANDR; medium effort; web search/fetch on frozen index, programmatic tools, code execution; 980k-token budget; five-hundred-task mean; Claude Opus 4.6 grader.

WANDR · 500-task wide research benchmark · Soft F1 · 允许工具 / medium / 框架未记录 · WANDR; medium effort; web search/fetch on frozen index, programmatic tools, code execution; 980k-token budget; five-hundred-task mean; Claude Opus 4.6 grader.
模型与配置成绩(%)记录与证据
Claude Opus 5.5Medium62.8%

允许工具 / high / 框架未记录 · WANDR; high effort; web search/fetch on frozen index, programmatic tools, code execution; 980k-token budget; five-hundred-task mean; Claude Opus 4.6 grader.

WANDR · 500-task wide research benchmark · Soft F1 · 允许工具 / high / 框架未记录 · WANDR; high effort; web search/fetch on frozen index, programmatic tools, code execution; 980k-token budget; five-hundred-task mean; Claude Opus 4.6 grader.
模型与配置成绩(%)记录与证据
Claude Opus 5.5High67.3%

允许工具 / xhigh / 框架未记录 · WANDR; xhigh effort; web search/fetch on frozen index, programmatic tools, code execution; 980k-token budget; five-hundred-task mean; Claude Opus 4.6 grader.

WANDR · 500-task wide research benchmark · Soft F1 · 允许工具 / xhigh / 框架未记录 · WANDR; xhigh effort; web search/fetch on frozen index, programmatic tools, code execution; 980k-token budget; five-hundred-task mean; Claude Opus 4.6 grader.
模型与配置成绩(%)记录与证据
Claude Opus 5.5Xhigh71.3%

允许工具 / max / 框架未记录 · WANDR; max effort; web search/fetch on frozen index, programmatic tools, code execution; 980k-token budget; five-hundred-task mean; Claude Opus 4.6 grader.

WANDR · 500-task wide research benchmark · Soft F1 · 允许工具 / max / 框架未记录 · WANDR; max effort; web search/fetch on frozen index, programmatic tools, code execution; 980k-token budget; five-hundred-task mean; Claude Opus 4.6 grader.
模型与配置成绩(%)记录与证据
Claude Opus 5.5Max72.3%

历史记录 · 4 次发布

  • Claude Fable 5.2本次发布有 5 条成绩记录,含 5 条精确值(可含历史记录)
  • Claude Haiku 5.5本次发布有 5 条成绩记录,含 5 条精确值(可含历史记录)
  • Claude Sonnet 5.5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 5.5本次发布有 5 条成绩记录,含 5 条精确值(可含历史记录)