Skip to content

Benchmark

HELM Safety: SimpleSafetyTests - 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
1.44

Score distribution

96.4%
Mean
98.8%
Median
8.6%
Std dev
92.3%
10th percentile
100%
90th percentile
43.3%
Min
43.3%100%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.44 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
1Claude 3 Haiku· Stanford HELM specialized suites data docs100%
2Claude 3 Opus· Stanford HELM specialized suites data docs100%
3Claude 3 Sonnet· Stanford HELM specialized suites data docs100%
4Claude 3.5 Sonnet· Stanford HELM specialized suites data docs100%
5Claude Opus 4thinking, 10k· Stanford HELM specialized suites data docs100%
6Claude Sonnet 3.7· Stanford HELM specialized suites data docs100%
7Claude Sonnet 4thinking, 10k· Stanford HELM specialized suites data docs100%
8Claude Sonnet 4.5· Stanford HELM specialized suites data docs100%
9Command R Plus· Stanford HELM specialized suites data docs100%
10GPT 4.1· Stanford HELM specialized suites data docs100%
11GPT OSS 120B· Stanford HELM specialized suites data docs100%
12GPT OSS 20b· Stanford HELM specialized suites data docs100%
13GPT-4.1 mini· Stanford HELM specialized suites data docs100%
14GPT-4.5 Preview· Stanford HELM specialized suites data docs100%
15GPT-5 mini· Stanford HELM specialized suites data docs100%
16GPT-5 nano· Stanford HELM specialized suites data docs100%
17Kimi K2 Instruct· Stanford HELM specialized suites data docs100%
18o4 Mini· Stanford HELM specialized suites data docs100%
19Palmyra X5· Stanford HELM specialized suites data docs100%
20Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs100%
21Qwen3 235B A22B Instruct 2507· Stanford HELM specialized suites data docs100%
22GPT 5· Stanford HELM specialized suites data docs99.8%
23GPT 5.1· Stanford HELM specialized suites data docs99.8%
24Claude Opus 4setting not named by the source· Stanford HELM specialized suites data docs99.5%
25Claude Sonnet 4setting not named by the source· Stanford HELM specialized suites data docs99.5%
26Qwen3 Next 80B A3Bthinking· Stanford HELM specialized suites data docs99.5%
27Grok 3 Mini· Stanford HELM specialized suites data docs99.3%
28Llama 4 Maverick· Stanford HELM specialized suites data docs99.3%
29Llama3 8B Chat· Stanford HELM specialized suites data docs99.3%
30GLM 4.5 Air· Stanford HELM specialized suites data docs99%
31GPT-4 Turbo· Stanford HELM specialized suites data docs99%
32GPT-4.1 nano· Stanford HELM specialized suites data docs99%
33o1· Stanford HELM specialized suites data docs99%
34o3 Mini· Stanford HELM specialized suites data docs99%
35OpenAI o3· Stanford HELM specialized suites data docs99%
36Qwen1.5-72B-Chat· Stanford HELM specialized suites data docs99%
37Claude Haiku 4.5· Stanford HELM specialized suites data docs98.8%
38Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs98.8%
39Llama 3.1 8B Instruct Turbo· Stanford HELM specialized suites data docs98.8%
40GPT 4o· Stanford HELM specialized suites data docs98.5%
41Qwen2 72B Instruct· Stanford HELM specialized suites data docs98.5%
42Gemini 2.0 Flash· Stanford HELM specialized suites data docs98.5%
43DeepSeek R1setting not named by the source· Stanford HELM specialized suites data docs98.3%
44Granite 4.0 H Small· Stanford HELM specialized suites data docs98%
45Olmo 2 32B Instruct· Stanford HELM specialized suites data docs98%
46gemini-2.5-flash-preview-04-17· Stanford HELM specialized suites data docs98%
47GPT-4o mini· Stanford HELM specialized suites data docs97.8%
48Gemini 2.0 Flash-Lite Preview· Stanford HELM specialized suites data docs97.7%
49DeepSeek R1setting not named by the source· Stanford HELM specialized suites data docs97.5%
50Granite 3.3 8b Instruct· Stanford HELM specialized suites data docs97.5%

Showing 1–50 of 74 results.

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