Harvey Legal Agent Benchmark: leaderboard
Legal-agent benchmark for completing realistic legal workflows with all-pass grading, including held-out Harvey tasks and public legal-agent task sets.
Metric: All-Pass Task Success (system card). Source: github.com. Source: benchmarklist.com. Source: www-cdn.anthropic.com. Saturation forecast: Around January 2028. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 | 23.58 | #12 |
| 2 | Claude Fable 5.1 (Max) | 19.09 | #9 (Claude Fable 5.1) |
| 3 | Claude Mythos 5 | 16.91 | #2 |
| 4 | Claude Mythos Preview | 13.4 | #4 |
| 5 | Claude Fable 5 | 13.3 | #16 |
| 6 | Claude Opus 4.8 | 10.4 | #37 |
| 7 | Claude Sonnet 5 | 8.92 | #57 |
| 8 | Claude Sonnet 4.6 | 8 | #90 |
| 9 | Claude Opus 4.7 | 7.1 | #49 |
| 10 | Claude Opus 4.6 | 4.2 | #63 |
| 11 | GPT-5.5 | 2.1 | #31 |
| 12 | Gemini 3.5 Flash | 0.8 | #60 |
| 13 | Gemini 3.1 Pro (Preview) | 0 | #58 |
Interactive version: theaggregate.ai/benchmark?slug=harvey-legal-agent-benchmark · How It Works · Data refreshed daily, snapshot 2026-10-20.