Skip to content

Benchmark

HELM AIR-Bench: AIRBench 2024 - #9: Violent acts

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
1.86

Score distribution

77%
Mean
86.7%
Median
24.5%
Std dev
35%
10th percentile
98.3%
90th percentile
5%
Min
5%100%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.86 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 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs100%
3Claude Haiku 4.5· Stanford HELM specialized suites data docs100%
4GPT OSS 120B· Stanford HELM specialized suites data docs100%
5GPT OSS 20b· Stanford HELM specialized suites data docs100%
6OpenAI o3· Stanford HELM specialized suites data docs100%
7Claude 3 Opus· Stanford HELM specialized suites data docs98.3%
8Claude 3 Sonnet· Stanford HELM specialized suites data docs98.3%
9Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs98.3%
10GPT 5· Stanford HELM specialized suites data docs98.3%
11GPT 5.1· Stanford HELM specialized suites data docs98.3%
12GPT-5 nano· Stanford HELM specialized suites data docs98.3%
13Qwen3 Next 80B A3Bthinking· Stanford HELM specialized suites data docs98.3%
14Claude Sonnet 4.5· Stanford HELM specialized suites data docs97.5%
15GPT-5 mini· Stanford HELM specialized suites data docs96.7%
16Granite 3.3 8b Instruct· Stanford HELM specialized suites data docs96.7%
17Granite 4.0 H Small· Stanford HELM specialized suites data docs96.7%
18Llama3 8B Chat· Stanford HELM specialized suites data docs96.7%
19o1· Stanford HELM specialized suites data docs96.7%
20o4 Mini· Stanford HELM specialized suites data docs96.7%
21Claude Sonnet 3.7· Stanford HELM specialized suites data docs95.8%
22o3 Mini· Stanford HELM specialized suites data docs95%
23Palmyra X5· Stanford HELM specialized suites data docs95%
24Qwen3 235B A22B Instruct 2507· Stanford HELM specialized suites data docs95%
25Claude Sonnet 4· Stanford HELM specialized suites data docs94.2%
26Claude Opus 4· Stanford HELM specialized suites data docs93.3%
27GPT-4 Turbo· Stanford HELM specialized suites data docs93.3%
28Gemini 3 Pro Preview· Stanford HELM specialized suites data docs91.7%
29GPT-4.5 Preview· Stanford HELM specialized suites data docs91.7%
30Kimi K2 Instruct· Stanford HELM specialized suites data docs91.7%
31GPT-4.1 nano· Stanford HELM specialized suites data docs90%
32Gemini 2.0 Flash-Lite Preview· Stanford HELM specialized suites data docs88.3%
33Gemini 2.5 Pro Preview 03-25· Stanford HELM specialized suites data docs88.3%
34gemini-2.0-pro-exp-02-05· Stanford HELM specialized suites data docs88.3%
35Gemini 2.0 Flash· Stanford HELM specialized suites data docs86.7%
36Gemini 2.5 Flash-Lite· Stanford HELM specialized suites data docs86.7%
37gemini-2.5-flash-preview-04-17· Stanford HELM specialized suites data docs86.7%
38GPT-4.1 mini· Stanford HELM specialized suites data docs86.7%
39GPT-4o mini· Stanford HELM specialized suites data docs86.7%
40gemini-1.5-flash-002· Stanford HELM specialized suites data docs85%
41Llama 4 Maverick· Stanford HELM specialized suites data docs85%
42Qwen2 72B Instruct· Stanford HELM specialized suites data docs85%
43GPT 4.1· Stanford HELM specialized suites data docs83.3%
44GPT 4osetting not named by the source· Stanford HELM specialized suites data docs83.3%
45GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs83.3%
46Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs83.3%
47gemini-1.5-pro-002· Stanford HELM specialized suites data docs81.7%
48GLM 4.5 Air· Stanford HELM specialized suites data docs81.7%
49GPT 4· Stanford HELM specialized suites data docs81.7%
50Llama 3.1 8B Instruct Turbo· Stanford HELM specialized suites data docs81.7%

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