Legal Agent Benchmark Leaderboard | Phaseo

Legal Agent Benchmark

AI benchmark results and model performance

Type: percentageLegal

Recorded Results

5

Average Score

10.60%

Score Range

5.80% - 13.30%

Leading Model (lowest score)

5.80% - Claude Sonnet 5

Scores Over Time

Individual benchmark scores plotted by date.

Models Using This Benchmark

Organisation	Model	Reported	Top Score	Info	Self Reported	Source
Anthropic	Claude Sonnet 5	30 Jun 2026	5.80%	Harvey held-out set; all-pass rate; reported in Anthropic's system card	Yes	Source
Anthropic	Claude Opus 5	24 Jul 2026	11.70%	Held-out all-pass rate reported by Harvey; 94.1% mean criterion-pass rate	No	Source
Anthropic	Claude Fable 5	09 Jun 2026	13.30%	Higher of Mythos 5 / Fable 5 in Anthropic's launch table	Yes	Source
Anthropic	Claude Mythos 5	09 Jun 2026	13.30%	Higher of Mythos 5 / Fable 5 in Anthropic's launch table	Yes	Source