Skip to content

Benchmark

HELM Safety: XSTest - 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
1.41

Score distribution

94.8%
Mean
96%
Median
4.7%
Std dev
90.7%
10th percentile
98.2%
90th percentile
68.7%
Min
68.7%98.8%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.41 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
51Granite 4.0 H Small· Stanford HELM specialized suites data docs94.6%
52Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs94.5%
53GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs94.2%
54Mixtral 8x22B Instruct v0.1· Stanford HELM specialized suites data docs94.1%
55o3 Mini· Stanford HELM specialized suites data docs94.1%
56Command R+· Stanford HELM specialized suites data docs93.9%
57Command R Plus· Stanford HELM specialized suites data docs93.8%
58GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs93.8%
59Claude Haiku 4.5· Stanford HELM specialized suites data docs93.2%
60GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs93.2%
61Mixtral 8x7B Instruct v0.1· Stanford HELM specialized suites data docs93.1%
62Gemini 2.0 Flash-Lite Preview· Stanford HELM specialized suites data docs92.7%
63GPT OSS 120B· Stanford HELM specialized suites data docs92.7%
64Mistral (7B) Instruct v0.3· Stanford HELM specialized suites data docs92.6%
65Claude 3 Opus· Stanford HELM specialized suites data docs92.5%
66gemini-1.5-flash-001· Stanford HELM specialized suites data docs92.1%
67Granite 3.3 8b Instruct· Stanford HELM specialized suites data docs90.7%
68gemini-1.5-pro-001· Stanford HELM specialized suites data docs90.4%
69DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs88.9%
70GPT OSS 20b· Stanford HELM specialized suites data docs87.2%
71Claude 3 Sonnet· Stanford HELM specialized suites data docs85.8%
72Claude 3 Haiku· Stanford HELM specialized suites data docs85.3%
73DBRX-instruct· Stanford HELM specialized suites data docs77.4%
74Mistral (7B) Instruct v0.1· Stanford HELM specialized suites data docs68.7%

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