Skip to content

Benchmark

HELM AIR-Bench: AIRBench 2024 - #6: Advice in Heavily Regulated Industries

Safety benchmarks examine risks in model behaviour, including how models respond to unsafe requests. Scores are percentages. Higher is better.

Models measured
70
Scores recorded
75
Sources
1
IQ weight
0.17

Score distribution

10.9%
Mean
8.7%
Median
8.3%
Std dev
1.3%
10th percentile
22.7%
90th percentile
0%
Min
0%40.7%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.17 for chat models, measured over 70 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Claude Haiku 4.5· Stanford HELM specialized suites data docs40.7%
2GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs31.3%
3Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs29.3%
4GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs26%
5Claude Sonnet 4.5· Stanford HELM specialized suites data docs24.7%
6GPT OSS 20b· Stanford HELM specialized suites data docs24.7%
7GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs24.7%
8GPT OSS 120B· Stanford HELM specialized suites data docs22.7%
9GPT-5 nano· Stanford HELM specialized suites data docs22%
10o1· Stanford HELM specialized suites data docs22%
11o3 Mini· Stanford HELM specialized suites data docs20.7%
12DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs19.3%
13GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs18%
14Claude 3 Opus· Stanford HELM specialized suites data docs17.3%
15gemini-1.5-flash-002· Stanford HELM specialized suites data docs17.3%
16Claude 3 Haiku· Stanford HELM specialized suites data docs16.7%
17gemini-2.5-flash-preview-04-17· Stanford HELM specialized suites data docs16%
18GPT 4· Stanford HELM specialized suites data docs16%
19Granite 4.0 H Small· Stanford HELM specialized suites data docs16%
20Grok 4· Stanford HELM specialized suites data docs16%
21DBRX-instruct· Stanford HELM specialized suites data docs15.3%
22gemini-2.0-pro-exp-02-05· Stanford HELM specialized suites data docs14.7%
23Grok 3 Mini· Stanford HELM specialized suites data docs14.7%
24Palmyra X5· Stanford HELM specialized suites data docs14%
25Claude Sonnet 4· Stanford HELM specialized suites data docs13.3%
26Gemini 2.5 Pro Preview 03-25· Stanford HELM specialized suites data docs13.3%
27Mixtral 8x22B Instruct v0.1· Stanford HELM specialized suites data docs13.3%
28Yi 34B Chat· Stanford HELM specialized suites data docs12.7%
29Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs12%
30O1 Mini· Stanford HELM specialized suites data docs12%
31Gemini 3 Pro Preview· Stanford HELM specialized suites data docs11.3%
32Claude 3 Sonnet· Stanford HELM specialized suites data docs10.7%
33OpenAI o3· Stanford HELM specialized suites data docs10.7%
34Gemini 2.0 Flash· Stanford HELM specialized suites data docs10%
35Qwen2 72B Instruct· Stanford HELM specialized suites data docs10%
36Qwen1.5-72B-Chat· Stanford HELM specialized suites data docs9.3%
37Granite 3.3 8b Instruct· Stanford HELM specialized suites data docs8.7%
38Llama 4 Scout· Stanford HELM specialized suites data docs8.7%
39Gemini 2.5 Flash-Lite· Stanford HELM specialized suites data docs8%
40GPT 5· Stanford HELM specialized suites data docs8%
41Qwen3 Next 80B A3Bthinking· Stanford HELM specialized suites data docs8%
42Gemini 2.0 Flash-Lite Preview· Stanford HELM specialized suites data docs7.3%
43Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs7.3%
44Llama3 8B Chat· Stanford HELM specialized suites data docs7.3%
45Claude Sonnet 3.7· Stanford HELM specialized suites data docs6.7%
46GPT-5 mini· Stanford HELM specialized suites data docs6.7%
47Grok 3· Stanford HELM specialized suites data docs6.7%
48Mistral (7B) Instruct v0.3· Stanford HELM specialized suites data docs6.7%
49Claude Opus 4· Stanford HELM specialized suites data docs6%
50Command R+· Stanford HELM specialized suites data docs6%

Showing 1–50 of 75 results.

Data credits: Stanford HELM (Stanford CRFM); scenario scores from Capabilities, Safety, Long Context, MedHELM, AIR-Bench, MMLU and Finance