PLawBench — leaderboard
Polish legal benchmark for practical legal-work evaluation, with legal questions and detailed rubric-based assessment.
Metric: Overall (self-reported). Source: benchmarklist.com. Status: saturation imminent. 24 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.2 | 69.67 |
| 2 | GPT-5 | 67.76 |
| 3 | Claude Opus 4.5 | 66.47 |
| 4 | Gemini 3 Pro (Preview) | 66.35 |
| 5 | Claude Sonnet 4.5 | 65.88 |
| 6 | Gemini 2.5 Pro | 64.05 |
| 7 | Qwen 3 235B A22B 2507 Instruct | 63.08 |
| 8 | Gemini 2.5 Flash | 62.07 |
| 9 | GLM-4.6 | 60.49 |
| 10 | DeepSeek V3.2 | 57.97 |
| 11 | Qwen 3 30B A3B 2507 Instruct | 55.73 |
| 12 | Claude Sonnet 4 | 53.55 |
| 13 | Grok 4.1 Fast | 52.67 |
| 14 | ERNIE 5.0 Thinking Preview | 51.39 |
| 15 | Qwen 3 4B 2507 Instruct | 50.53 |
Interactive version: theaggregate.ai/benchmark?slug=plawbench · How the rankings work · Data refreshed daily, snapshot 2026-07-22.