LegalBench-RU: leaderboard
Russian legal benchmark over lawyer-authored cases: overall accuracy % across three tracks — statute and rule recall, reasoning over facts and court positions, and MCP tool calls against state registries.
Metric: Overall Accuracy (%). Source: huggingface.co. Saturation forecast: Around January 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V3.2 | 76 | #205 |
| 2 | Qwen 3 235B A22B | 71 | #313 |
| 3 | GPT-OSS-120B | 57 | #341 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=legalbench-ru · How It Works · Data refreshed daily, snapshot 2026-10-20.