TW-LegalBench: leaderboard
Large language models (LLMs) have shown impressive capabilities across diverse tasks, yet their performance on jurisdiction-specific legal reasoning remains underexplored. We present TW-LegalBench that utilizes Taiwanese legal system's rich official corpus open to the public to fill the gap in evaluating LLMs on Taiwanese law, among common-law benchmarks that focus on English sources and civil-law
Metric: Total (self-reported). Source: benchmarklist.com. Saturation forecast: Around February 2027. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.5 | 81 | #142 |
| 2 | GPT-5 | 76.7 | #98 |
| 3 | Qwen 3 235B A22B | 69.1 | #313 |
| 4 | GPT-4o | 66.2 | #343 |
| 5 | Llama 3.1 405B | 60.5 | #466 |
| 6 | GPT-OSS-120B | 56 | #341 |
| 7 | nemotron-3-nano-30B-a3B | 52.6 | #662 |
| 8 | Qwen 2.5 7B | 49.7 | #922 |
| 9 | GPT-OSS-20B | 48.3 | #518 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=tw-legalbench · How It Works · Data refreshed daily, snapshot 2026-10-20.