Skip to content

Benchmark

HELM AIR-Bench: AIRBench 2024 - Defamation

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
1.04

Score distribution

68.1%
Mean
64.8%
Median
21%
Std dev
38.9%
10th percentile
97.2%
90th percentile
22.2%
Min
22.2%100%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.04 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
1Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs100%
2Claude Sonnet 4.5· Stanford HELM specialized suites data docs100%
3Claude Opus 4· Stanford HELM specialized suites data docs99.1%
4Claude Sonnet 4· Stanford HELM specialized suites data docs99.1%
5Claude Haiku 4.5· Stanford HELM specialized suites data docs98.1%
6GPT 5.1· Stanford HELM specialized suites data docs98.1%
7GPT OSS 20b· Stanford HELM specialized suites data docs98.1%
8GPT 5· Stanford HELM specialized suites data docs97.2%
9Claude 3 Haiku· Stanford HELM specialized suites data docs96.3%
10Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs96.3%
11GPT OSS 120B· Stanford HELM specialized suites data docs96.3%
12GPT-5 nano· Stanford HELM specialized suites data docs95.4%
13Claude 3 Opus· Stanford HELM specialized suites data docs92.6%
14Claude 3 Sonnet· Stanford HELM specialized suites data docs91.7%
15Claude Sonnet 3.7· Stanford HELM specialized suites data docs91.7%
16Granite 3.3 8b Instruct· Stanford HELM specialized suites data docs91.7%
17GPT-5 mini· Stanford HELM specialized suites data docs90.7%
18Qwen3 Next 80B A3Bthinking· Stanford HELM specialized suites data docs89.8%
19o1· Stanford HELM specialized suites data docs88%
20OpenAI o3· Stanford HELM specialized suites data docs88%
21Palmyra X5· Stanford HELM specialized suites data docs88%
22GPT-4 Turbo· Stanford HELM specialized suites data docs85.2%
23GPT-4.5 Preview· Stanford HELM specialized suites data docs85.2%
24Granite 4.0 H Small· Stanford HELM specialized suites data docs84.3%
25Kimi K2 Instruct· Stanford HELM specialized suites data docs80.6%
26o4 Mini· Stanford HELM specialized suites data docs80.6%
27o3 Mini· Stanford HELM specialized suites data docs77.8%
28Qwen3 235B A22B Instruct 2507· Stanford HELM specialized suites data docs77.8%
29Gemini 2.5 Pro Preview 03-25· Stanford HELM specialized suites data docs73.1%
30Gemini 3 Pro Preview· Stanford HELM specialized suites data docs73.1%
31GPT-4.1 nano· Stanford HELM specialized suites data docs71.3%
32Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs71.3%
33GPT 4.1· Stanford HELM specialized suites data docs69.4%
34GPT 4osetting not named by the source· Stanford HELM specialized suites data docs69.4%
35Llama3 8B Chat· Stanford HELM specialized suites data docs68.5%
36Llama 3.1 8B Instruct Turbo· Stanford HELM specialized suites data docs66.7%
37gemini-2.5-flash-preview-04-17· Stanford HELM specialized suites data docs65.7%
38Gemini 2.5 Flash-Lite· Stanford HELM specialized suites data docs64.8%
39GLM 4.5 Air· Stanford HELM specialized suites data docs64.8%
40Llama 4 Maverick· Stanford HELM specialized suites data docs63.9%
41Qwen2 72B Instruct· Stanford HELM specialized suites data docs63.9%
42GPT-4.1 mini· Stanford HELM specialized suites data docs62%
43Gemini 2.0 Flash-Lite Preview· Stanford HELM specialized suites data docs61.1%
44DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs60.2%
45GPT 4osetting not named by the source· Stanford HELM specialized suites data docs60.2%
46GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs60.2%
47Grok 3 Mini· Stanford HELM specialized suites data docs60.2%
48Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs60.2%
49DeepSeek R1· Stanford HELM specialized suites data docs59.3%
50Gemini 2.0 Flash· Stanford HELM specialized suites data docs58.3%

Showing 1–50 of 75 results.

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