Skip to content

Benchmark

HELM AIR-Bench: AIRBench 2024 - #21: Endangerment, Harm, or Abuse of Children

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
1.01

Score distribution

51.3%
Mean
50.7%
Median
23.1%
Std dev
18.7%
10th percentile
82.7%
90th percentile
6.7%
Min
6.7%89.3%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.01 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
51Llama 3.1 8B Instruct Turbo· Stanford HELM specialized suites data docs40%
52GPT 4osetting not named by the source· Stanford HELM specialized suites data docs38%
53Grok 3 Mini· Stanford HELM specialized suites data docs37.3%
54GPT 4osetting not named by the source· Stanford HELM specialized suites data docs36.7%
55Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs36.7%
56GPT-4o mini· Stanford HELM specialized suites data docs35.3%
57O1 Mini· Stanford HELM specialized suites data docs35.3%
58DeepSeek R1· Stanford HELM specialized suites data docs33.3%
59Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs30%
60Yi 34B Chat· Stanford HELM specialized suites data docs28.7%
61DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs27.3%
62GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs25.3%
63Grok 3· Stanford HELM specialized suites data docs25.3%
64DeepSeek V3· Stanford HELM specialized suites data docs21.3%
65Qwen1.5-72B-Chat· Stanford HELM specialized suites data docs20.7%
66Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs20.7%
67Mistral Large2407· Stanford HELM specialized suites data docs20%
68Mixtral 8x22B Instruct v0.1· Stanford HELM specialized suites data docs18.7%
69GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs17.3%
70Mixtral 8x7B Instruct v0.1· Stanford HELM specialized suites data docs13.3%
71Command R+· Stanford HELM specialized suites data docs12.7%
72Command R Plus· Stanford HELM specialized suites data docs11.3%
73Mistral Small· Stanford HELM specialized suites data docs10.7%
74DBRX-instruct· Stanford HELM specialized suites data docs9.3%
75Mistral (7B) Instruct v0.3· Stanford HELM specialized suites data docs6.7%

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