Skip to content

Benchmark

HELM AIR-Bench: AIRBench 2024 - #11: Weapon Usage & Development

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
1.28

Score distribution

68.3%
Mean
75.6%
Median
24.8%
Std dev
31.1%
10th percentile
95.6%
90th percentile
5.6%
Min
5.6%100%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.28 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
51GPT 4osetting not named by the source· Stanford HELM specialized suites data docs60%
52Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs60%
53Grok 3· Stanford HELM specialized suites data docs59.4%
54Grok 4· Stanford HELM specialized suites data docs55.6%
55Qwen1.5-72B-Chat· Stanford HELM specialized suites data docs55.6%
56DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs54.4%
57GLM 4.5 Air· Stanford HELM specialized suites data docs50%
58Llama 4 Scout· Stanford HELM specialized suites data docs48.3%
59Grok 3 Mini· Stanford HELM specialized suites data docs47.8%
60Yi 34B Chat· Stanford HELM specialized suites data docs47.8%
61GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs47.2%
62Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs37.8%
63GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs36.7%
64O1 Mini· Stanford HELM specialized suites data docs36.7%
65DeepSeek R1· Stanford HELM specialized suites data docs34.4%
66DeepSeek V3· Stanford HELM specialized suites data docs32.2%
67Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs32.2%
68Mistral Large2407· Stanford HELM specialized suites data docs31.1%
69Mixtral 8x7B Instruct v0.1· Stanford HELM specialized suites data docs26.7%
70Mixtral 8x22B Instruct v0.1· Stanford HELM specialized suites data docs25.6%
71Command R+· Stanford HELM specialized suites data docs21.1%
72Mistral Small· Stanford HELM specialized suites data docs20%
73Command R Plus· Stanford HELM specialized suites data docs11.1%
74Mistral (7B) Instruct v0.3· Stanford HELM specialized suites data docs7.8%
75DBRX-instruct· Stanford HELM specialized suites data docs5.6%

Showing 51–75 of 75 results.

Data credits: Stanford HELM (Stanford CRFM); scenario scores from Capabilities, Safety, Long Context, MedHELM, AIR-Bench, MMLU and Finance