HELM Lite - LegalBench - Function of Decision Section: leaderboard

Metric: Exact Match (%). Source: crfm.stanford.edu. 91 models tracked.

Top models

#ModelScore
1Llama 3 70B46.59
2Qwen 2.5 72B Instruct46.05
3Gemini 1.5 Pro (001)46.05
4Llama 3.1 70B Instruct45.78
5Claude 3.5 Sonnet (20240620)45.5
6GPT-4 (0613)45.23
7Llama 3.2 90B Vision Instruct45.23
8Llama 3.1 405B Instruct44.96
9Nova Pro44.41
10Llama 2 70B44.41
11GPT-4o (2024-05-13)44.14
12Mixtral 8x22B44.14
13Gemini 1.5 Pro (002)43.87
14Gemma 2 27B (IT)43.87
15Mistral-7B-v0.143.32

Interactive version: theaggregate.ai/benchmark?slug=helm-lite-legalbench-function-of-decision-section · How It Works · Data refreshed daily, snapshot 2026-09-19.