Skip to content

Benchmark

HELM AIR-Bench: AIRBench 2024 - Privacy

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
1.06

Score distribution

69.3%
Mean
69.4%
Median
18.9%
Std dev
43.3%
10th percentile
91.6%
90th percentile
19.4%
Min
19.4%96.5%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.06 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-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs61.8%
52Grok 3· Stanford HELM specialized suites data docs61.1%
53Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs60.8%
54Yi 34B Chat· Stanford HELM specialized suites data docs59.6%
55Llama 4 Scout· Stanford HELM specialized suites data docs58.3%
56Grok 3 Mini· Stanford HELM specialized suites data docs57.1%
57GPT 4osetting not named by the source· Stanford HELM specialized suites data docs56%
58DeepSeek R1· Stanford HELM specialized suites data docs55.5%
59GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs53.4%
60GLM 4.5 Air· Stanford HELM specialized suites data docs50%
61DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs49.6%
62Grok 4· Stanford HELM specialized suites data docs49.6%
63Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs47.5%
64Qwen1.5-72B-Chat· Stanford HELM specialized suites data docs46.9%
65Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs44.9%
66Mixtral 8x7B Instruct v0.1· Stanford HELM specialized suites data docs44.8%
67Mixtral 8x22B Instruct v0.1· Stanford HELM specialized suites data docs43.5%
68O1 Mini· Stanford HELM specialized suites data docs43.3%
69DeepSeek V3· Stanford HELM specialized suites data docs41.1%
70Mistral Large2407· Stanford HELM specialized suites data docs36.6%
71Mistral Small· Stanford HELM specialized suites data docs35.3%
72Command R Plus· Stanford HELM specialized suites data docs31.1%
73Command R+· Stanford HELM specialized suites data docs31%
74Mistral (7B) Instruct v0.3· Stanford HELM specialized suites data docs30.2%
75DBRX-instruct· Stanford HELM specialized suites data docs19.4%

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