Skip to content

Benchmark

HELM AIR-Bench: AIRBench 2024 - #32: Fraud

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
0.98

Score distribution

55.6%
Mean
50.7%
Median
26.5%
Std dev
20%
10th percentile
92%
90th percentile
6.7%
Min
6.7%100%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.98 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 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs100%
2Claude Haiku 4.5· Stanford HELM specialized suites data docs97.3%
3Claude Sonnet 4.5· Stanford HELM specialized suites data docs97.3%
4GPT 5· Stanford HELM specialized suites data docs93.3%
5GPT OSS 120B· Stanford HELM specialized suites data docs93.3%
6GPT OSS 20b· Stanford HELM specialized suites data docs93.3%
7Claude Opus 4· Stanford HELM specialized suites data docs92.7%
8Claude Sonnet 4· Stanford HELM specialized suites data docs92%
9GPT-5 nano· Stanford HELM specialized suites data docs91.3%
10OpenAI o3· Stanford HELM specialized suites data docs91.3%
11Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs90.7%
12GPT-5 mini· Stanford HELM specialized suites data docs90.7%
13Claude 3 Haiku· Stanford HELM specialized suites data docs89.3%
14GPT 5.1· Stanford HELM specialized suites data docs89.3%
15Claude 3 Sonnet· Stanford HELM specialized suites data docs88%
16Qwen3 Next 80B A3Bthinking· Stanford HELM specialized suites data docs84%
17Kimi K2 Instruct· Stanford HELM specialized suites data docs82.7%
18Granite 3.3 8b Instruct· Stanford HELM specialized suites data docs81.3%
19Claude 3 Opus· Stanford HELM specialized suites data docs80%
20Palmyra X5· Stanford HELM specialized suites data docs78.7%
21Qwen3 235B A22B Instruct 2507· Stanford HELM specialized suites data docs78.7%
22Gemini 3 Pro Preview· Stanford HELM specialized suites data docs76%
23o1· Stanford HELM specialized suites data docs75.3%
24o4 Mini· Stanford HELM specialized suites data docs74.7%
25Claude Sonnet 3.7· Stanford HELM specialized suites data docs72%
26GPT-4.5 Preview· Stanford HELM specialized suites data docs72%
27Llama 4 Maverick· Stanford HELM specialized suites data docs71.3%
28Granite 4.0 H Small· Stanford HELM specialized suites data docs66.7%
29o3 Mini· Stanford HELM specialized suites data docs66.7%
30Llama3 8B Chat· Stanford HELM specialized suites data docs66%
31GPT 4.1· Stanford HELM specialized suites data docs62.7%
32Llama 3.1 8B Instruct Turbo· Stanford HELM specialized suites data docs60.7%
33GPT-4 Turbo· Stanford HELM specialized suites data docs60%
34Gemini 2.5 Pro Preview 03-25· Stanford HELM specialized suites data docs58.7%
35gemini-2.0-pro-exp-02-05· Stanford HELM specialized suites data docs56%
36Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs53.3%
37Gemini 2.0 Flash· Stanford HELM specialized suites data docs51.3%
38Gemini 2.0 Flash-Lite Preview· Stanford HELM specialized suites data docs50.7%
39GPT 4· Stanford HELM specialized suites data docs50.7%
40GPT-4.1 mini· Stanford HELM specialized suites data docs50.7%
41Llama 4 Scout· Stanford HELM specialized suites data docs49.3%
42Qwen2 72B Instruct· Stanford HELM specialized suites data docs49.3%
43Gemini 2.5 Flash-Lite· Stanford HELM specialized suites data docs48.7%
44GPT 4osetting not named by the source· Stanford HELM specialized suites data docs46.7%
45GPT-4.1 nano· Stanford HELM specialized suites data docs45.3%
46gemini-2.5-flash-preview-04-17· Stanford HELM specialized suites data docs43.3%
47Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs42.7%
48GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs42%
49gemini-1.5-pro-002· Stanford HELM specialized suites data docs41.3%
50GPT 4osetting not named by the source· Stanford HELM specialized suites data docs41.3%

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