XpertBench - Law: leaderboard
Metric: Weighted rubric score (%) on the XpertBench-Gold subset (245 expert-written open-ended professional tasks, 15-40 weighted binary checkpoints each); ShotJudge: Gemini 2.5 Pro judges every checkpoint with an expert-graded GPT-5 response as one-shot exemplar; law tasks; higher is better. Source: arxiv.org. Saturation forecast: Around April 2028. 14 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.6 (Thinking) | 65.54 |
| 2 | GPT-5.4 (High) | 64.79 |
| 3 | Claude Sonnet 4.5 (Thinking) | 58.7 |
| 4 | Kimi K2 (Thinking) | 58 |
| 5 | GPT-5.2 (High) | 56.88 |
| 6 | GPT-5 (High) | 54.7 |
| 7 | Gemini 3.1 Pro (Preview) | 53.88 |
| 8 | DeepSeek V3.1 Terminus (Thinking) | 51.9 |
| 9 | Gemini 3 Flash | 50.22 |
| 10 | Doubao-Seed-1.6 (Thinking) | 49.4 |
| 11 | Gemini 3 Pro | 48.59 |
| 12 | Gemini 2.5 Pro | 47.3 |
| 13 | GLM-4.6 | 47.2 |
Interactive version: theaggregate.ai/benchmark?slug=xpertbench-law · How It Works · Data refreshed daily, snapshot 2026-10-07.