Skip to content

Benchmark

HELM AIR-Bench: AIRBench 2024 - Fundamental Rights

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
1.68

Score distribution

77.9%
Mean
86.7%
Median
24.3%
Std dev
40%
10th percentile
100%
90th percentile
5.3%
Min
5.3%100%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.68 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
51gemini-1.5-pro-002· Stanford HELM specialized suites data docs76%
52gemini-1.5-flash-002· Stanford HELM specialized suites data docs73.3%
53GPT 4osetting not named by the source· Stanford HELM specialized suites data docs73.3%
54Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs73.3%
55GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs69.3%
56DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs64%
57Llama 4 Scout· Stanford HELM specialized suites data docs63.3%
58Grok 3 Mini· Stanford HELM specialized suites data docs62.7%
59Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs62.7%
60DeepSeek R1· Stanford HELM specialized suites data docs60%
61Grok 3· Stanford HELM specialized suites data docs57.3%
62Qwen1.5-72B-Chat· Stanford HELM specialized suites data docs53.3%
63Mixtral 8x22B Instruct v0.1· Stanford HELM specialized suites data docs48%
64DeepSeek V3· Stanford HELM specialized suites data docs46.7%
65GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs46.7%
66Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs45.3%
67O1 Mini· Stanford HELM specialized suites data docs41.3%
68Grok 4· Stanford HELM specialized suites data docs40%
69Mixtral 8x7B Instruct v0.1· Stanford HELM specialized suites data docs37.3%
70Mistral Large2407· Stanford HELM specialized suites data docs32.7%
71Mistral Small· Stanford HELM specialized suites data docs29.3%
72Command R+· Stanford HELM specialized suites data docs22.7%
73Command R Plus· Stanford HELM specialized suites data docs20%
74Mistral (7B) Instruct v0.3· Stanford HELM specialized suites data docs12%
75DBRX-instruct· Stanford HELM specialized suites data docs5.3%

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