LexRubric: leaderboard
Metric: Overall score rate (%): points earned from expert-written atomic rubric items (weighted -10 to +10; negative items subtract) divided by the maximum positive score, averaged over all 649 instances (473 legal consultation, 176 judicial examination); a Qwen3.6-27B judge checks each item, temperature 0.6, 16k output tokens; can fall below 0; higher is better. Source: arxiv.org. Saturation forecast: Around September 2027. 18 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Kimi K2.6 | 75.21 |
| 2 | Qwen 3.6 Max Preview | 75.12 |
| 3 | Qwen 3 Max | 74.35 |
| 4 | Kimi K2.5 | 73.4 |
| 5 | GLM-5.1 | 69.96 |
| 6 | Claude Sonnet 4.6 | 69.18 |
| 7 | Qwen 3.5 397B A17B | 68.03 |
| 8 | GLM-5 | 63.28 |
| 9 | DeepSeek V4 Flash | 63.19 |
| 10 | DeepSeek V4 Pro | 62.22 |
| 11 | DeepSeek V3.2 | 60.03 |
| 12 | DeepSeek R1 | 59.49 |
| 13 | GPT-5.2 | 56.43 |
Interactive version: theaggregate.ai/benchmark?slug=lexrubric · How It Works · Data refreshed daily, snapshot 2026-09-29.