Skip to content

Benchmark

HELM Safety: Anthropic Red Team - LM Evaluated Safety score

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
1.07

Score distribution

97.7%
Mean
99%
Median
4.4%
Std dev
95.7%
10th percentile
99.8%
90th percentile
71.1%
Min
71.1%100%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.07 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
51OpenAI o3· Stanford HELM specialized suites data docs98.3%
52o4 Mini· Stanford HELM specialized suites data docs98.2%
53Claude Sonnet 4.5· Stanford HELM specialized suites data docs98.2%
54Llama 4 Maverick· Stanford HELM specialized suites data docs98.2%
55Command R Plus· Stanford HELM specialized suites data docs98%
56GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs98%
57DeepSeek V3· Stanford HELM specialized suites data docs97.1%
58Gemini 3 Pro Preview· Stanford HELM specialized suites data docs97.1%
59Claude Haiku 4.5· Stanford HELM specialized suites data docs96.9%
60Llama 3.1 8B Instruct Turbo· Stanford HELM specialized suites data docs96.8%
61GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs96.7%
62Llama 4 Scout· Stanford HELM specialized suites data docs96.5%
63Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs96.5%
64Mistral Small· Stanford HELM specialized suites data docs96.4%
65DeepSeek R1setting not named by the source· Stanford HELM specialized suites data docs96.2%
66Mixtral 8x22B Instruct v0.1· Stanford HELM specialized suites data docs96.2%
67Grok 4· Stanford HELM specialized suites data docs95.7%
68Grok 3· Stanford HELM specialized suites data docs95.5%
69Command R+· Stanford HELM specialized suites data docs93.7%
70Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs93.2%
71Mixtral 8x7B Instruct v0.1· Stanford HELM specialized suites data docs92.8%
72Mistral (7B) Instruct v0.3· Stanford HELM specialized suites data docs90.1%
73DBRX-instruct· Stanford HELM specialized suites data docs76.6%
74Mistral (7B) Instruct v0.1· Stanford HELM specialized suites data docs71.1%

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