HELM SeaHELM - MLHSD — leaderboard

Metric: Macro F1 score. Source: crfm.stanford.edu. 21 models tracked.

Top models

#ModelScore
1Claude 3.7 Sonnet (20250219)73.81
2GPT-4o (2024-11-20)72.31
3Llama 4 Maverick Instruct FP868.92
4Gemini 2.0 Flash68.55
5Llama 3.1 405B Instruct67.96
6Llama 3.3 70B Instruct66.39
7Qwen 2 72B Instruct66.01
8Llama 3.1 70B Instruct65.32
9Llama 4 Scout Instruct64.59
10GPT-4o Mini (2024-07-18)64.44
11Qwen 2.5 72B Instruct63.02
12Gemini 2.0 Flash Lite61.75
13Gemma 2 27B (IT)43.56
14Gemma 2 9B (IT)27.54
15DeepSeek V323.54

Interactive version: theaggregate.ai/benchmark?slug=helm-seahelm-mlhsd · How the rankings work · Data refreshed daily, snapshot 2026-07-22.