Skip to content

Benchmark

HELM AIR-Bench: AIRBench 2024 - #3: Availability

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
1.26

Score distribution

71.1%
Mean
77.8%
Median
25.3%
Std dev
33.3%
10th percentile
100%
90th percentile
2.8%
Min
2.8%100%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.26 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 docs100%
3Claude Sonnet 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%
6GPT-5 mini· Stanford HELM specialized suites data docs100%
7GPT-5 nano· Stanford HELM specialized suites data docs100%
8o4 Mini· Stanford HELM specialized suites data docs100%
9Claude 3 Haiku· Stanford HELM specialized suites data docs97.2%
10Claude 3 Sonnet· Stanford HELM specialized suites data docs97.2%
11OpenAI o3· Stanford HELM specialized suites data docs97.2%
12Claude 3 Opus· Stanford HELM specialized suites data docs95.8%
13Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs95.8%
14Claude Sonnet 4· Stanford HELM specialized suites data docs94.4%
15Llama3 8B Chat· Stanford HELM specialized suites data docs94.4%
16GPT 5· Stanford HELM specialized suites data docs91.7%
17GPT 5.1· Stanford HELM specialized suites data docs91.7%
18Kimi K2 Instruct· Stanford HELM specialized suites data docs91.7%
19o1· Stanford HELM specialized suites data docs91.7%
20Qwen3 Next 80B A3Bthinking· Stanford HELM specialized suites data docs91.7%
21Claude Opus 4· Stanford HELM specialized suites data docs90.3%
22Granite 3.3 8b Instruct· Stanford HELM specialized suites data docs90.3%
23Gemini 2.0 Flash-Lite Preview· Stanford HELM specialized suites data docs88.9%
24GPT 4· Stanford HELM specialized suites data docs88.9%
25GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs88.9%
26o3 Mini· Stanford HELM specialized suites data docs88.9%
27GPT-4 Turbo· Stanford HELM specialized suites data docs86.1%
28GPT-4.5 Preview· Stanford HELM specialized suites data docs83.3%
29GPT-4o mini· Stanford HELM specialized suites data docs83.3%
30Granite 4.0 H Small· Stanford HELM specialized suites data docs83.3%
31Palmyra X5· Stanford HELM specialized suites data docs83.3%
32Gemini 2.0 Flash· Stanford HELM specialized suites data docs80.6%
33GPT 4osetting not named by the source· Stanford HELM specialized suites data docs80.6%
34Gemini 2.5 Flash-Lite· Stanford HELM specialized suites data docs77.8%
35Gemini 2.5 Pro Preview 03-25· Stanford HELM specialized suites data docs77.8%
36Gemini 3 Pro Preview· Stanford HELM specialized suites data docs77.8%
37GPT 4.1· Stanford HELM specialized suites data docs77.8%
38Qwen3 235B A22B Instruct 2507· Stanford HELM specialized suites data docs77.8%
39Claude Sonnet 3.7· Stanford HELM specialized suites data docs76.4%
40gemini-2.0-pro-exp-02-05· Stanford HELM specialized suites data docs75%
41gemini-2.5-flash-preview-04-17· Stanford HELM specialized suites data docs75%
42GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs75%
43GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs75%
44GPT-4.1 mini· Stanford HELM specialized suites data docs75%
45GPT-4.1 nano· Stanford HELM specialized suites data docs75%
46gemini-1.5-flash-002· Stanford HELM specialized suites data docs72.2%
47Llama 4 Maverick· Stanford HELM specialized suites data docs72.2%
48GPT 4osetting not named by the source· Stanford HELM specialized suites data docs69.4%
49Llama 3.1 8B Instruct Turbo· Stanford HELM specialized suites data docs69.4%
50Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs66.7%

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