PLawBench: leaderboard
Rubric-based benchmark for real-world legal practice: 850 questions across 13 practical legal scenarios, each scored against expert-written rubrics.
Metric: Overall (self-reported). Source: benchmarklist.com. Status: years away from saturation. 29 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3.8 Max | 73.2 |
| 2 | GPT-5.6 Sol (Max) | 72.3 |
| 3 | Claude Fable 5 | 70.2 |
| 4 | GPT-5.2 | 69.67 |
| 5 | Claude Opus 4.8 | 69.6 |
| 6 | GPT-5 | 67.76 |
| 7 | Claude Opus 4.5 | 66.47 |
| 8 | Gemini 3 Pro (Preview) | 66.35 |
| 9 | Claude Sonnet 4.5 | 65.88 |
| 10 | Qwen 3 Max | 64.75 |
| 11 | Gemini 2.5 Pro | 64.05 |
| 12 | Qwen 3 235B A22B 2507 Instruct | 63.08 |
| 13 | Gemini 2.5 Flash | 62.07 |
| 14 | GLM-4.6 | 60.49 |
| 15 | Qwen 3.7 Max | 58.9 |
Interactive version: theaggregate.ai/benchmark?slug=plawbench · How It Works · Data refreshed daily, snapshot 2026-09-05.