Skip to content

Benchmark

HELM Safety: BBQ - BBQ accuracy

Safety benchmarks examine risks in model behaviour, including how models respond to unsafe requests. Scores are percentages. Higher is better.

Models measured
69
Scores recorded
74
Sources
1
IQ weight
1.95

Score distribution

90.2%
Mean
94.1%
Median
9.9%
Std dev
72.4%
10th percentile
97.7%
90th percentile
56.5%
Min
56.5%99.3%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.95 for chat models, measured over 69 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
51Qwen3 Next 80B A3Bthinking· Stanford HELM specialized suites data docs91.7%
52Mixtral 8x22B Instruct v0.1· Stanford HELM specialized suites data docs91.6%
53Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs90.6%
54Claude 3 Sonnet· Stanford HELM specialized suites data docs90%
55Command R Plus· Stanford HELM specialized suites data docs89.9%
56GPT 5.1· Stanford HELM specialized suites data docs88.7%
57GPT-4o mini· Stanford HELM specialized suites data docs88.2%
58GPT-4.1 nano· Stanford HELM specialized suites data docs87.5%
59Llama 4 Scout· Stanford HELM specialized suites data docs87.5%
60DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs86.2%
61Mixtral 8x7B Instruct v0.1· Stanford HELM specialized suites data docs85.7%
62Qwen1.5-72B-Chat· Stanford HELM specialized suites data docs84.6%
63Granite 3.3 8b Instruct· Stanford HELM specialized suites data docs84%
64DBRX-instruct· Stanford HELM specialized suites data docs79.2%
65Llama 3.1 8B Instruct Turbo· Stanford HELM specialized suites data docs78.5%
66Llama3 8B Chat· Stanford HELM specialized suites data docs76.5%
67Command R+· Stanford HELM specialized suites data docs72.4%
68Olmo 2 32B Instruct· Stanford HELM specialized suites data docs71.4%
69GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs69.9%
70Mistral (7B) Instruct v0.3· Stanford HELM specialized suites data docs66.8%
71GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs64.9%
72Claude 3 Haiku· Stanford HELM specialized suites data docs62.5%
73GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs60.6%
74Mistral (7B) Instruct v0.1· Stanford HELM specialized suites data docs56.5%

Showing 51–74 of 74 results.

Data credits: Stanford HELM (Stanford CRFM); scenario scores from Capabilities, Safety, Long Context, MedHELM, AIR-Bench, MMLU and Finance