LegalBench - Rhetoric Tasks
Professional Work · 2023-08-18
The Rhetoric Tasks split of LegalBench. This child benchmark separates a source-reported subtask or subtrack from the parent aggregate so scores at different grains do not share one benchmark_slug.
Top models (higher is better)
| Model | Score |
|---|---|
| Gemini 3.1 Pro Preview | 84.1 |
| Gemini 3 Pro Preview | 84.1 |
| Gemini 3 Flash Preview | 83.9 |
| Claude Fable 5 | 83.8 |
| Muse Wonder Full | 83.6 |
| Gemini 3.5 Flash | 83.4 |
| GPT-5.1 | 82.9 |
| Grok 4.6 | 82.9 |
| GPT-5.2 | 82.9 |
| Gemini 4 Argon | 82.8 |
| GPT-5.4 | 82.8 |
| Grok 4.5 | 82.8 |
| GLM-5.3 | 82.8 |
| GPT-5.6 Luna | 82.7 |
| Grok 4 | 82.7 |
| Gemini 3.7 Flash | 82.5 |
| MiniMax M3 | 82.5 |
| GPT-5.5 | 82.4 |
| Claude Fable 5.1 | 82.3 |
| Gemini 2.5 Flash Preview 04-17 | 82.2 |
| Muse Spark 1.2 | 82.2 |
| Muse Spark 1.1 | 82.1 |
| GPT-5.6 Terra | 82.1 |
| Gemini 2.5 Pro Experimental 03-25 | 82.1 |
| GPT-5.6 Sol | 82.1 |
| Kimi K3 | 81.9 |
| Gemini 3.6 Flash | 81.7 |
| Qwen3.7-Max (source unspecified) | 81.7 |
| Hy4 Preview | 81.6 |
| Grok 3 Mini | 81.6 |
| Grok 3 (API) | 81.6 |
| o3 | 81.5 |
| Grok 4.7 | 81.5 |
| Opus 4.6 | 81.5 |
| Qwen3 Max (rolling alias) | 81.4 |
| Gemini 2.5 Flash Preview (09-2025) | 81.4 |
| Kimi K2.6 | 81.4 |
| Qwen3.8 27B | 81.4 |
| Gemini 3.8 Flash | 81.3 |
| Grok 4 Fast | 81.3 |