Skip to content

Benchmark

HELM AIR-Bench: AIRBench 2024 - #33: Academic dishonesty

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
1.17

Score distribution

73.8%
Mean
80%
Median
25.8%
Std dev
33.3%
10th percentile
100%
90th percentile
10%
Min
10%100%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.17 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 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 Haiku 4.5· Stanford HELM specialized suites data docs100%
6Claude Sonnet 4.5· Stanford HELM specialized suites data docs100%
7GPT 5· Stanford HELM specialized suites data docs100%
8GPT 5.1· Stanford HELM specialized suites data docs100%
9GPT OSS 120B· Stanford HELM specialized suites data docs100%
10GPT OSS 20b· Stanford HELM specialized suites data docs100%
11GPT-5 nano· Stanford HELM specialized suites data docs100%
12Llama3 8B Chat· Stanford HELM specialized suites data docs100%
13o4 Mini· Stanford HELM specialized suites data docs100%
14OpenAI o3· Stanford HELM specialized suites data docs100%
15Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs96.7%
16Claude Sonnet 3.7· Stanford HELM specialized suites data docs96.7%
17Claude Sonnet 4· Stanford HELM specialized suites data docs96.7%
18GPT-5 mini· Stanford HELM specialized suites data docs96.7%
19Granite 3.3 8b Instruct· Stanford HELM specialized suites data docs96.7%
20o1· Stanford HELM specialized suites data docs96.7%
21o3 Mini· Stanford HELM specialized suites data docs96.7%
22Qwen3 235B A22B Instruct 2507· Stanford HELM specialized suites data docs96.7%
23Qwen3 Next 80B A3Bthinking· Stanford HELM specialized suites data docs96.7%
24Claude Opus 4· Stanford HELM specialized suites data docs93.3%
25Gemini 2.5 Pro Preview 03-25· Stanford HELM specialized suites data docs93.3%
26Gemini 3 Pro Preview· Stanford HELM specialized suites data docs93.3%
27GPT-4 Turbo· Stanford HELM specialized suites data docs93.3%
28GPT-4.5 Preview· Stanford HELM specialized suites data docs93.3%
29Kimi K2 Instruct· Stanford HELM specialized suites data docs93.3%
30Llama 3.1 8B Instruct Turbo· Stanford HELM specialized suites data docs93.3%
31Palmyra X5· Stanford HELM specialized suites data docs90%
32Granite 4.0 H Small· Stanford HELM specialized suites data docs86.7%
33Qwen2 72B Instruct· Stanford HELM specialized suites data docs86.7%
34GPT 4.1· Stanford HELM specialized suites data docs83.3%
35gemini-1.5-flash-002· Stanford HELM specialized suites data docs80%
36gemini-1.5-pro-002· Stanford HELM specialized suites data docs80%
37GLM 4.5 Air· Stanford HELM specialized suites data docs80%
38GPT-4.1 mini· Stanford HELM specialized suites data docs80%
39Llama 4 Maverick· Stanford HELM specialized suites data docs80%
40Gemini 2.0 Flash· Stanford HELM specialized suites data docs76.7%
41Gemini 2.0 Flash-Lite Preview· Stanford HELM specialized suites data docs76.7%
42gemini-2.0-pro-exp-02-05· Stanford HELM specialized suites data docs76.7%
43GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs76.7%
44GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs76.7%
45GPT-4.1 nano· Stanford HELM specialized suites data docs76.7%
46Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs76.7%
47GPT 4· Stanford HELM specialized suites data docs73.3%
48GPT 4osetting not named by the source· Stanford HELM specialized suites data docs73.3%
49GPT-4o mini· Stanford HELM specialized suites data docs73.3%
50Gemini 2.5 Flash-Lite· Stanford HELM specialized suites data docs71.7%

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