LegalBench - Conclusion Tasks
Professional Work · 2023-08-18
The Conclusion 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 |
|---|---|
| Claude Fable 5.1 | 92.5 |
| Claude Fable 5 | 91.8 |
| Claude Opus 5 | 91.6 |
| Gemini 3.7 Flash | 91.1 |
| Gemini 3.6 Flash | 91.1 |
| GPT-5 | 90.9 |
| Gemini 4 Argon | 90.8 |
| Gemini 3.8 Flash | 90.7 |
| GPT-5.6 Sol | 90.5 |
| Grok 4.6 | 90.4 |
| Gemini 3 Pro Preview | 90.4 |
| GPT-5.6 Terra | 90.3 |
| Muse Wonder Full | 90.3 |
| GLM-5.3 | 90.3 |
| GPT-5.2 | 90.2 |
| GLM-4.7 | 90.1 |
| Gemini 3 Flash Preview | 90.1 |
| Kimi K3 | 90.1 |
| Kimi K2 Thinking | 90.0 |
| Opus 4.7 | 90.0 |
| Qwen3.5 Plus (2026-02-15) | 89.9 |
| Grok 4.3 | 89.9 |
| Qwen3.7-Max (source unspecified) | 89.8 |
| GPT-5.5 | 89.6 |
| GLM-5.1 | 89.6 |
| GLM-5.3 Flash | 89.6 |
| Sonnet 5 | 89.6 |
| DeepSeek V4 Pro 0813 | 89.5 |
| Grok 4.5 | 89.5 |
| Muse Spark | 89.5 |
| Opus 4.5 | 89.4 |
| GPT-5.1 | 89.4 |
| Kimi K2.6 | 89.4 |
| MiniMax M2.7 | 89.3 |
| MiniMax M3 | 89.3 |
| Gemini 3.1 Pro Preview | 89.3 |
| Qwen3.6 Plus (2026-04-02) | 89.2 |
| DeepSeek-V4-Pro | 89.1 |
| DeepSeek V4.1 Flash | 89.1 |
| Inkling | 89.0 |