DRACO · 100-task Perplexity deep-research benchmark · 测试定义与成绩

DRACO · 100-task Perplexity deep-research benchmark 收录 8 次模型发布的评测披露或成绩。

测试版本与指标

别名 Deep Research Accuracy, Completeness, and Objectivity

  • Normalized score · % · 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。

Normalized score %

允许工具 / low / 框架未记录 · DRACO 100-task set; low effort; web search/fetch, programmatic tool calling, code execution; 980k-token budget.

DRACO · 100-task Perplexity deep-research benchmark · Normalized score · 允许工具 / low / 框架未记录 · DRACO 100-task set; low effort; web search/fetch, programmatic tool calling, code execution; 980k-token budget.
模型与配置成绩(%)记录与证据
Claude Fable 5.2Low81.6%

允许工具 / medium / 框架未记录 · DRACO 100-task set; medium effort; web search/fetch, programmatic tool calling, code execution; 980k-token budget.

DRACO · 100-task Perplexity deep-research benchmark · Normalized score · 允许工具 / medium / 框架未记录 · DRACO 100-task set; medium effort; web search/fetch, programmatic tool calling, code execution; 980k-token budget.
模型与配置成绩(%)记录与证据
Claude Fable 5.2Medium82.2%

允许工具 / high / 框架未记录 · DRACO 100-task set; high effort; web search/fetch, programmatic tool calling, code execution; 980k-token budget.

DRACO · 100-task Perplexity deep-research benchmark · Normalized score · 允许工具 / high / 框架未记录 · DRACO 100-task set; high effort; web search/fetch, programmatic tool calling, code execution; 980k-token budget.
模型与配置成绩(%)记录与证据
Claude Fable 5.2High84.1%

允许工具 / xhigh / 框架未记录 · DRACO 100-task set; xhigh effort; web search/fetch, programmatic tool calling, code execution; 980k-token budget.

DRACO · 100-task Perplexity deep-research benchmark · Normalized score · 允许工具 / xhigh / 框架未记录 · DRACO 100-task set; xhigh effort; web search/fetch, programmatic tool calling, code execution; 980k-token budget.
模型与配置成绩(%)记录与证据
Claude Fable 5.2Xhigh85.2%

允许工具 / max / 框架未记录 · DRACO 100-task set; max effort; web search/fetch, programmatic tool calling, code execution; 980k-token budget.

DRACO · 100-task Perplexity deep-research benchmark · Normalized score · 允许工具 / max / 框架未记录 · DRACO 100-task set; max effort; web search/fetch, programmatic tool calling, code execution; 980k-token budget.
模型与配置成绩(%)记录与证据
Claude Fable 5.2Max87.7%

允许工具 / low / 框架未记录 · DRACO; 100 curated deep-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; five independent grader runs per response, averaged.

DRACO · 100-task Perplexity deep-research benchmark · Normalized score · 允许工具 / low / 框架未记录 · DRACO; 100 curated deep-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; five independent grader runs per response, averaged.
模型与配置成绩(%)记录与证据
Claude Haiku 5.5Low64.3%

允许工具 / medium / 框架未记录 · DRACO; 100 curated deep-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; five independent grader runs per response, averaged.

DRACO · 100-task Perplexity deep-research benchmark · Normalized score · 允许工具 / medium / 框架未记录 · DRACO; 100 curated deep-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; five independent grader runs per response, averaged.
模型与配置成绩(%)记录与证据
Claude Haiku 5.5Medium72.4%

允许工具 / high / 框架未记录 · DRACO; 100 curated deep-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; five independent grader runs per response, averaged.

DRACO · 100-task Perplexity deep-research benchmark · Normalized score · 允许工具 / high / 框架未记录 · DRACO; 100 curated deep-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; five independent grader runs per response, averaged.
模型与配置成绩(%)记录与证据
Claude Haiku 5.5High77.8%

允许工具 / xhigh / 框架未记录 · DRACO; 100 curated deep-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; five independent grader runs per response, averaged.

DRACO · 100-task Perplexity deep-research benchmark · Normalized score · 允许工具 / xhigh / 框架未记录 · DRACO; 100 curated deep-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; five independent grader runs per response, averaged.
模型与配置成绩(%)记录与证据
Claude Haiku 5.5Xhigh80.5%

允许工具 / max / 框架未记录 · DRACO; 100 curated deep-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; five independent grader runs per response, averaged.

DRACO · 100-task Perplexity deep-research benchmark · Normalized score · 允许工具 / max / 框架未记录 · DRACO; 100 curated deep-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; five independent grader runs per response, averaged.
模型与配置成绩(%)记录与证据
Claude Haiku 5.5Max81.5%

允许工具 / low / 框架未记录 · DRACO, 100 curated tasks; normalized score from five independent rubric-grading runs per response.

DRACO · 100-task Perplexity deep-research benchmark · Normalized score · 允许工具 / low / 框架未记录 · DRACO, 100 curated tasks; normalized score from five independent rubric-grading runs per response.
模型与配置成绩(%)记录与证据
Claude Sonnet 5.5Low64.6%

允许工具 / med / 框架未记录 · DRACO, 100 curated tasks; normalized score from five independent rubric-grading runs per response.

DRACO · 100-task Perplexity deep-research benchmark · Normalized score · 允许工具 / med / 框架未记录 · DRACO, 100 curated tasks; normalized score from five independent rubric-grading runs per response.
模型与配置成绩(%)记录与证据
Claude Sonnet 5.5Medium71.2%

允许工具 / high / 框架未记录 · DRACO, 100 curated tasks; normalized score from five independent rubric-grading runs per response.

DRACO · 100-task Perplexity deep-research benchmark · Normalized score · 允许工具 / high / 框架未记录 · DRACO, 100 curated tasks; normalized score from five independent rubric-grading runs per response.
模型与配置成绩(%)记录与证据
Claude Sonnet 5.5High81%

允许工具 / xhigh / 框架未记录 · DRACO, 100 curated tasks; normalized score from five independent rubric-grading runs per response.

DRACO · 100-task Perplexity deep-research benchmark · Normalized score · 允许工具 / xhigh / 框架未记录 · DRACO, 100 curated tasks; normalized score from five independent rubric-grading runs per response.
模型与配置成绩(%)记录与证据
Claude Sonnet 5.5Xhigh85.7%

允许工具 / max / 框架未记录 · DRACO, 100 curated tasks; normalized score from five independent rubric-grading runs per response.

DRACO · 100-task Perplexity deep-research benchmark · Normalized score · 允许工具 / max / 框架未记录 · DRACO, 100 curated tasks; normalized score from five independent rubric-grading runs per response.
模型与配置成绩(%)记录与证据
Claude Sonnet 5.5Max87%

允许工具 / low / 框架未记录 · DRACO; low effort; web search/fetch, programmatic tools, code execution; 980k-token budget; own agent harness; five judge grading runs per response, mean; Claude Opus 4.6 grader.

DRACO · 100-task Perplexity deep-research benchmark · Normalized score · 允许工具 / low / 框架未记录 · DRACO; low effort; web search/fetch, programmatic tools, code execution; 980k-token budget; own agent harness; five judge grading runs per response, mean; Claude Opus 4.6 grader.
模型与配置成绩(%)记录与证据
Claude Opus 5.5Low72.5%

允许工具 / medium / 框架未记录 · DRACO; medium effort; web search/fetch, programmatic tools, code execution; 980k-token budget; own agent harness; five judge grading runs per response, mean; Claude Opus 4.6 grader.

DRACO · 100-task Perplexity deep-research benchmark · Normalized score · 允许工具 / medium / 框架未记录 · DRACO; medium effort; web search/fetch, programmatic tools, code execution; 980k-token budget; own agent harness; five judge grading runs per response, mean; Claude Opus 4.6 grader.
模型与配置成绩(%)记录与证据
Claude Opus 5.5Medium83.9%

允许工具 / high / 框架未记录 · DRACO; high effort; web search/fetch, programmatic tools, code execution; 980k-token budget; own agent harness; five judge grading runs per response, mean; Claude Opus 4.6 grader.

DRACO · 100-task Perplexity deep-research benchmark · Normalized score · 允许工具 / high / 框架未记录 · DRACO; high effort; web search/fetch, programmatic tools, code execution; 980k-token budget; own agent harness; five judge grading runs per response, mean; Claude Opus 4.6 grader.
模型与配置成绩(%)记录与证据
Claude Opus 5.5High85%

允许工具 / xhigh / 框架未记录 · DRACO; xhigh effort; web search/fetch, programmatic tools, code execution; 980k-token budget; own agent harness; five judge grading runs per response, mean; Claude Opus 4.6 grader.

DRACO · 100-task Perplexity deep-research benchmark · Normalized score · 允许工具 / xhigh / 框架未记录 · DRACO; xhigh effort; web search/fetch, programmatic tools, code execution; 980k-token budget; own agent harness; five judge grading runs per response, mean; Claude Opus 4.6 grader.
模型与配置成绩(%)记录与证据
Claude Opus 5.5Xhigh86.7%

允许工具 / max / 框架未记录 · DRACO; max effort; web search/fetch, programmatic tools, code execution; 980k-token budget; own agent harness; five judge grading runs per response, mean; Claude Opus 4.6 grader.

DRACO · 100-task Perplexity deep-research benchmark · Normalized score · 允许工具 / max / 框架未记录 · DRACO; max effort; web search/fetch, programmatic tools, code execution; 980k-token budget; own agent harness; five judge grading runs per response, mean; Claude Opus 4.6 grader.
模型与配置成绩(%)记录与证据
Claude Opus 5.5Max87.4%

历史记录 · 8 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)
  • Claude Fable 5.1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 4.8本次发布有 1 条成绩记录,含 1 条精确值(可含历史记录)
  • Claude Opus 4.7本次发布有 1 条成绩记录,含 1 条精确值(可含历史记录)