Multi-task Indic Language Understanding Benchmark · 10 Indic languages plus English · overall average · 测试定义与成绩

Multi-task Indic Language Understanding Benchmark · 10 Indic languages plus English · overall average 收录 9 次模型发布的评测披露或成绩。

测试版本与指标

别名 MILU、Multi-task Indic Language Understanding Benchmark

  • Average accuracy · % · 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。

Average accuracy %

禁止工具 / max / 框架未记录 · MILU average accuracy across 11 languages; max effort; five-trial average; no tools or custom system prompt.

Multi-task Indic Language Understanding Benchmark · 10 Indic languages plus English · overall average · Average accuracy · 禁止工具 / max / 框架未记录 · MILU average accuracy across 11 languages; max effort; five-trial average; no tools or custom system prompt.
模型与配置成绩(%)记录与证据
Claude Fable 5.2Max94.1%

禁止工具 / max / 框架未记录 · MILU; average accuracy over 11 languages; no tools or customized system prompt; mean of five trials.

Multi-task Indic Language Understanding Benchmark · 10 Indic languages plus English · overall average · Average accuracy · 禁止工具 / max / 框架未记录 · MILU; average accuracy over 11 languages; no tools or customized system prompt; mean of five trials.
模型与配置成绩(%)记录与证据
Claude Haiku 5.5Max87.6%

禁止工具 / max / 框架未记录 · MILU average accuracy across 10 Indic languages plus English; five-trial average.

Multi-task Indic Language Understanding Benchmark · 10 Indic languages plus English · overall average · Average accuracy · 禁止工具 / max / 框架未记录 · MILU average accuracy across 10 Indic languages plus English; five-trial average.
模型与配置成绩(%)记录与证据
Claude Sonnet 5.5Max91.6%

禁止工具 / max / 框架未记录 · MILU average accuracy across 11 languages; adaptive thinking/max effort; five trials; no tools or customized system prompt.

Multi-task Indic Language Understanding Benchmark · 10 Indic languages plus English · overall average · Average accuracy · 禁止工具 / max / 框架未记录 · MILU average accuracy across 11 languages; adaptive thinking/max effort; five trials; no tools or customized system prompt.
模型与配置成绩(%)记录与证据
Claude Opus 5.5Max93.1%

历史记录 · 9 次发布

  • Claude Fable 5.2本次发布有 1 条成绩记录,含 1 条精确值(可含历史记录)
  • Claude Haiku 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)
  • Claude Sonnet 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)
  • 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)
  • Claude Opus 5本次发布有 1 条成绩记录,含 1 条精确值(可含历史记录)
  • Claude Sonnet 5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 4.8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)
  • Claude Sonnet 4.6本次发布有 1 条成绩记录,含 1 条精确值(可含历史记录)