Legal Agent Benchmark (BenchmarkList Mixed Public and Held-Out): leaderboard
BenchmarkList imports all-pass legal workflow results from model system cards. Its main column combines results reported on public legal-agent tasks and Harvey held-out tasks. The imported cohort preserves each reported score and remains separate from the primary public and held-out populations.
Metric: All-Pass Task Success (system card). Source: benchmarklist.com. Saturation forecast: Around February 2028. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 | 23.58 | #13 |
| 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 | #3 |
| 5 | Claude Fable 5 | 13.3 | #17 |
| 6 | Claude Opus 4.8 | 10.4 | #36 |
| 7 | Claude Sonnet 5 | 8.92 | #60 |
| 8 | Claude Sonnet 4.6 | 8 | #93 |
| 9 | Claude Opus 4.7 | 7.1 | #51 |
| 10 | Claude Opus 4.6 | 4.2 | #64 |
| 11 | GPT-5.5 | 2.1 | #30 |
| 12 | Gemini 3.5 Flash | 0.8 | #62 |
| 13 | Gemini 3.1 Pro (Preview) | 0 | #59 |
No result here: #4 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=legal-agent-benchmark-benchmarklist-mixed-public-and-held-out · How It Works · Data refreshed daily, snapshot 2026-10-24.