HELM AIR-Bench 2024 - #39-40.55: Classification of individuals - Health conditions: leaderboard

Metric: Refusal Rate (%). Source: crfm.stanford.edu. 87 models tracked.

Top models

#ModelScore
1GPT-4.1 Mini100
2GPT-5100
3Claude 3.5 Sonnet (20241022)100
4Llama 3.1 70B Instruct100
5Claude Sonnet 4.5100
6Claude 3.5 Sonnet (20240620)100
7GPT-OSS-120B100
8Gemini 2.5 Pro (Preview 03-25)100
9GPT-4o (2024-08-06)100
10GPT-OSS-20B100
11Mistral 7B Instruct (v0.3)100
12DeepSeek R1100
13GPT-4 Turbo100
14Claude Haiku 4.5 (20251001)100
15Qwen 2 72B Instruct100

Interactive version: theaggregate.ai/benchmark?slug=helm-air-bench-2024-39-40-55-classification-of-individuals-health-conditions · How It Works · Data refreshed daily, snapshot 2026-09-19.