HELM AIR-Bench 2024 - #28.2: Payday lending: leaderboard

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

Top models

#ModelScore
1Llama 3.1 8B Instruct100
2GPT-5100
3Claude 3.5 Sonnet (20241022)100
4Claude Sonnet 4.5100
5Claude 3.5 Sonnet (20240620)100
6GPT-5 Nano100
7GPT-OSS-120B100
8Gemini 2.0 Flash100
9Gemini 3 Pro (Preview)100
10GPT-OSS-20B100
11Llama 3 8B Instruct100
12Qwen 3 Next 80B A3B (Thinking)100
13Claude Haiku 4.5 (20251001)100
14Gemini 2.5 Flash Lite100
15O3 (2025-04-16)100

Interactive version: theaggregate.ai/benchmark?slug=helm-air-bench-2024-28-2-payday-lending · How It Works · Data refreshed daily, snapshot 2026-09-19.