Legal Benchmarks - Data Extraction - Criteria Pass Rate: leaderboard
Legal Benchmarks Data Extraction: share of applicable lawyer-authored pass/fail criteria satisfied in both of a model's two attempts, graded one criterion at a time by a calibrated GLM 5.3 Flash judge. Shows coverage of the requested terms even when a whole task fails; %, higher is better. Tables are replaced each monthly cycle.
Metric: Criteria pass rate (%, criteria met in both attempts). Source: www.legalbenchmarks.ai. Saturation forecast: Around February 2027. 26 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5.5 | 91.1 | #3 |
| 2 | Claude Fable 5.1 | 89.9 | #10 |
| 3 | Claude Opus 5 | 89.2 | #16 |
| 4 | GPT-6 Astra | 88.4 | #6 |
| 5 | Muse Spark 1.3 | 86.5 | #31 |
| 6 | Gemini 4 Argon | 86.2 | #9 |
| 7 | Qwen 3.8 Flash | 85.7 | #62 |
| 8 | GPT-6.1 Sol | 84.5 | #7 |
| 9 | Claude Sonnet 5.5 | 83.3 | #14 |
| 10 | Claude Haiku 5.5 | 82.3 | #43 |
| 11 | GPT-5.6 Sol | 82 | #22 |
| 12 | Kimi K3 | 81.8 | #35 |
| 13 | MiniMax-M3 | 79.8 | #151 |
| 14 | Claude Sonnet 5 | 78.8 | #65 |
| 15 | Grok 4.6 | 78.3 | #29 |
Interactive version: theaggregate.ai/benchmark?slug=legal-benchmarks-data-extraction-criteria-pass-rate · How It Works · Data refreshed daily, snapshot 2026-11-05.