Skip to content

Benchmark

HELM AIR-Bench: AIRBench 2024 - #16: Offensive Language

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
0.77

Score distribution

61.2%
Mean
61.9%
Median
21.4%
Std dev
29.8%
10th percentile
90.5%
90th percentile
16.7%
Min
16.7%97.6%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.77 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
51Grok 3· Stanford HELM specialized suites data docs48.8%
52Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs48.8%
53DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs47.6%
54O1 Mini· Stanford HELM specialized suites data docs47.6%
55Qwen1.5-72B-Chat· Stanford HELM specialized suites data docs47.6%
56Mixtral 8x22B Instruct v0.1· Stanford HELM specialized suites data docs46.4%
57Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs46.4%
58DeepSeek V3· Stanford HELM specialized suites data docs45.2%
59GPT-4o mini· Stanford HELM specialized suites data docs45.2%
60Llama 4 Scout· Stanford HELM specialized suites data docs45.2%
61OpenAI o3· Stanford HELM specialized suites data docs45.2%
62Grok 3 Mini· Stanford HELM specialized suites data docs40.5%
63Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs39.3%
64o4 Mini· Stanford HELM specialized suites data docs39.3%
65o1· Stanford HELM specialized suites data docs36.9%
66Llama 3.1 8B Instruct Turbo· Stanford HELM specialized suites data docs35.7%
67Mistral (7B) Instruct v0.3· Stanford HELM specialized suites data docs31%
68Mistral Large2407· Stanford HELM specialized suites data docs29.8%
69Command R Plus· Stanford HELM specialized suites data docs27.4%
70Command R+· Stanford HELM specialized suites data docs26.2%
71DBRX-instruct· Stanford HELM specialized suites data docs26.2%
72Grok 4· Stanford HELM specialized suites data docs22.6%
73Mixtral 8x7B Instruct v0.1· Stanford HELM specialized suites data docs21.4%
74Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs19%
75Mistral Small· Stanford HELM specialized suites data docs16.7%

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