Polish Legal RAG - Legal Reasoning: leaderboard
Polish Legal RAG leaderboard by gaius-lex, sum_rag_v2 task: advanced Polish legal reasoning over provided sources, covering contradiction resolution and whether hypothetical situations meet legal criteria. Score 0-100, higher is better; 16 models, results frozen since August 2025.
Metric: Score (0-100). Source: huggingface.co. Saturation forecast: Around February 2027. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GLM 4.5 Air | 87.72 | #403 |
| 2 | GLM-4.5 | 85.68 | #283 |
| 3 | Qwen 3 30B A3B | 84.19 | #496 |
| 4 | Qwen 3 14B | 83.09 | #542 |
| 5 | Qwen 3 32B | 78.99 | #449 |
| 6 | GPT-OSS-120B | 77.37 | #346 |
| 7 | Kimi K2 | 76.48 | #252 |
| 8 | Qwen 3 235B A22B 2507 Instruct | 75.02 | #315 |
| 9 | Qwen 3 235B A22B 2507 (Thinking) | 73.67 | #263 |
| 10 | PLLuM-12B-nc-chat | 65.23 | #1109 |
| 11 | GPT-OSS-20B | 54.49 | #523 |
| 12 | Llama 3 8B Instruct | 53.03 | #1116 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=polish-legal-rag-legal-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-05.