Harvey LAB-AA: leaderboard
Artificial Analysis evaluates legal work across practice areas using Harvey LAB. Rubric coverage and the rate of tasks passing every criterion are separate percentages; neither is an Elo rating.
Metric: Hallucination-Gated All-Pass Rate (%). Source: artificialanalysis.ai. Saturation forecast: Around 2030. 28 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Muse Wonder Full (Max) | 13.06 | #12 (Muse Wonder Full) |
| 2 | Muse Wonder (Max) | 10 | #18 (Muse Wonder) |
| 3 | Grok 4.7 (xHigh) | 9.44 | #26 (Grok 4.7) |
| 4 | Muse Spark 1.3 (Max) | 8.89 | #31 (Muse Spark 1.3) |
| 5 | GPT-6 Astra (Max) | 8.61 | #6 (GPT-6 Astra) |
| 6 | GPT-6.1 Sol (Max) | 6.94 | #7 (GPT-6.1 Sol) |
| 7 | Kimi K3 (Max) | 5.28 | #35 (Kimi K3) |
| 8 | Gemini 4 Argon (High) | 3.89 | #9 (Gemini 4 Argon) |
| 9 | GPT-6 Sol (Max) | 3.61 | #19 (GPT-6 Sol) |
| 10 | GPT-6 Luna (Max) | 3.33 | #61 (GPT-6 Luna) |
| 11 | Mistral Large 4 | 3.06 | #106 |
| 12 | Claude Haiku 5.5 (xHigh) | 2.78 | #43 (Claude Haiku 5.5) |
| 13 | DeepSeek V4.1 Flash (Max) | 1.67 | #37 (DeepSeek V4.1 Flash) |
| 14 | Claude Haiku 5.5 (Low) | 1.67 | #43 (Claude Haiku 5.5) |
| 15 | Claude Haiku 5.5 (High) | 1.11 | #43 (Claude Haiku 5.5) |
No result here: #3 Claude Opus 5.5, #10 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=harvey-lab-aa · How It Works · Data refreshed daily, snapshot 2026-11-05.