Skip to content

Benchmark

HELM AIR-Bench: AIRBench 2024 - #23: Suicidal and non-suicidal self Injury

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
1.99

Score distribution

83.4%
Mean
90%
Median
17.6%
Std dev
56.7%
10th percentile
100%
90th percentile
27.8%
Min
27.8%100%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.99 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 docs81.1%
52GPT-4.1 mini· Stanford HELM specialized suites data docs80%
53GPT-4o mini· Stanford HELM specialized suites data docs78.9%
54Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs78.9%
55Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs77.8%
56DeepSeek R1· Stanford HELM specialized suites data docs75.6%
57Grok 3· Stanford HELM specialized suites data docs73.3%
58Yi 34B Chat· Stanford HELM specialized suites data docs71.1%
59Mixtral 8x22B Instruct v0.1· Stanford HELM specialized suites data docs70%
60Llama 4 Scout· Stanford HELM specialized suites data docs68.9%
61GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs67.8%
62DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs65.6%
63Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs65.6%
64Mistral Large2407· Stanford HELM specialized suites data docs64.4%
65Qwen1.5-72B-Chat· Stanford HELM specialized suites data docs63.3%
66DeepSeek V3· Stanford HELM specialized suites data docs62.2%
67Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs62.2%
68GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs56.7%
69O1 Mini· Stanford HELM specialized suites data docs51.1%
70Mixtral 8x7B Instruct v0.1· Stanford HELM specialized suites data docs45.6%
71Mistral (7B) Instruct v0.3· Stanford HELM specialized suites data docs44.4%
72Command R Plus· Stanford HELM specialized suites data docs43.3%
73Mistral Small· Stanford HELM specialized suites data docs42.2%
74Command R+· Stanford HELM specialized suites data docs41.1%
75DBRX-instruct· Stanford HELM specialized suites data docs27.8%

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