Skip to content

Benchmark

HELM AIR-Bench: AIRBench 2024 - #43: Illegal/Regulated substances/goods

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
1.68

Score distribution

85.9%
Mean
93.3%
Median
21.4%
Std dev
46.7%
10th percentile
100%
90th percentile
13.3%
Min
13.3%100%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.68 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 3 Haiku· Stanford HELM specialized suites data docs100%
2Claude 3 Opus· Stanford HELM specialized suites data docs100%
3Claude 3 Sonnet· Stanford HELM specialized suites data docs100%
4Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs100%
5Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs100%
6Claude Haiku 4.5· Stanford HELM specialized suites data docs100%
7Claude Sonnet 3.7· Stanford HELM specialized suites data docs100%
8Claude Sonnet 4· Stanford HELM specialized suites data docs100%
9Claude Sonnet 4.5· Stanford HELM specialized suites data docs100%
10Gemini 2.0 Flash· Stanford HELM specialized suites data docs100%
11Gemini 2.0 Flash-Lite Preview· Stanford HELM specialized suites data docs100%
12Gemini 2.5 Flash-Lite· Stanford HELM specialized suites data docs100%
13Gemini 2.5 Pro Preview 03-25· Stanford HELM specialized suites data docs100%
14gemini-2.0-pro-exp-02-05· Stanford HELM specialized suites data docs100%
15GPT 4.1· Stanford HELM specialized suites data docs100%
16GPT 5· Stanford HELM specialized suites data docs100%
17GPT 5.1· Stanford HELM specialized suites data docs100%
18GPT OSS 120B· Stanford HELM specialized suites data docs100%
19GPT OSS 20b· Stanford HELM specialized suites data docs100%
20GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs100%
21GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs100%
22GPT-4 Turbo· Stanford HELM specialized suites data docs100%
23GPT-4.5 Preview· Stanford HELM specialized suites data docs100%
24GPT-5 mini· Stanford HELM specialized suites data docs100%
25GPT-5 nano· Stanford HELM specialized suites data docs100%
26Granite 3.3 8b Instruct· Stanford HELM specialized suites data docs100%
27Granite 4.0 H Small· Stanford HELM specialized suites data docs100%
28Llama3 8B Chat· Stanford HELM specialized suites data docs100%
29o1· Stanford HELM specialized suites data docs100%
30o3 Mini· Stanford HELM specialized suites data docs100%
31o4 Mini· Stanford HELM specialized suites data docs100%
32OpenAI o3· Stanford HELM specialized suites data docs100%
33Palmyra X5· Stanford HELM specialized suites data docs100%
34Qwen3 235B A22B Instruct 2507· Stanford HELM specialized suites data docs100%
35Qwen3 Next 80B A3Bthinking· Stanford HELM specialized suites data docs100%
36Claude Opus 4· Stanford HELM specialized suites data docs93.3%
37Gemini 3 Pro Preview· Stanford HELM specialized suites data docs93.3%
38gemini-1.5-flash-002· Stanford HELM specialized suites data docs93.3%
39gemini-2.5-flash-preview-04-17· Stanford HELM specialized suites data docs93.3%
40GPT 4· Stanford HELM specialized suites data docs93.3%
41GPT-4.1 nano· Stanford HELM specialized suites data docs93.3%
42GPT-4o mini· Stanford HELM specialized suites data docs93.3%
43Llama 3.1 8B Instruct Turbo· Stanford HELM specialized suites data docs93.3%
44Llama 4 Maverick· Stanford HELM specialized suites data docs93.3%
45Llama 4 Scout· Stanford HELM specialized suites data docs93.3%
46Qwen2 72B Instruct· Stanford HELM specialized suites data docs93.3%
47DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs86.7%
48gemini-1.5-pro-002· Stanford HELM specialized suites data docs86.7%
49GLM 4.5 Air· Stanford HELM specialized suites data docs86.7%
50GPT 4osetting not named by the source· Stanford HELM specialized suites data docs86.7%

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