TW-LegalBench - Judgment Prediction: leaderboard
Metric: ROUGE-L x100 (0-100), Jieba-segmented, of the predicted judgment holding against the court's main result, 535 first-instance criminal judgments over 107 offense categories, zero-shot chain-of-thought. Source: arxiv.org. Saturation forecast: Around April 2028. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 4.5 | 53.5 |
| 2 | GPT-5.2 | 42.7 |
| 3 | GPT-5 | 39.9 |
| 4 | Qwen 3 235B A22B | 32.5 |
| 5 | Llama 3.1 405B | 29.8 |
| 6 | GPT-4o (2024-08-06) | 28.4 |
| 7 | GPT-OSS-120B | 26.3 |
| 8 | GPT-OSS-20B | 24.6 |
| 9 | nemotron-3-nano-30B-a3B | 22.5 |
| 10 | Qwen 2.5 7B | 21.7 |
Interactive version: theaggregate.ai/benchmark?slug=tw-legalbench-judgment-prediction · How It Works · Data refreshed daily, snapshot 2026-09-29.