Skip to content

Benchmark

HELM AIR-Bench: AIRBench 2024 - #14: Hate Speech (Inciting/Promoting/Expressing hatred)

Scores are percentages. Higher is better.

Models measured
70
Scores recorded
75
Sources
1
IQ weight
1.39

Score distribution

86.3%
Mean
90.6%
Median
12.7%
Std dev
67%
10th percentile
98.8%
90th percentile
50.4%
Min
50.4%100%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.39 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 Haiku 4.5· Stanford HELM specialized suites data docs100%
2Claude Sonnet 4· Stanford HELM specialized suites data docs99.9%
3Claude Opus 4· Stanford HELM specialized suites data docs99.7%
4Claude Sonnet 4.5· Stanford HELM specialized suites data docs99.7%
5Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs99.6%
6Claude Sonnet 3.7· Stanford HELM specialized suites data docs99%
7Qwen3 Next 80B A3Bthinking· Stanford HELM specialized suites data docs99%
8GPT 5.1· Stanford HELM specialized suites data docs98.8%
9Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs98.6%
10Qwen3 235B A22B Instruct 2507· Stanford HELM specialized suites data docs98.3%
11Claude 3 Opus· Stanford HELM specialized suites data docs98.1%
12Claude 3 Haiku· Stanford HELM specialized suites data docs98%
13GPT OSS 20b· Stanford HELM specialized suites data docs98%
14GPT 5· Stanford HELM specialized suites data docs97.8%
15Claude 3 Sonnet· Stanford HELM specialized suites data docs97.7%
16GPT OSS 120B· Stanford HELM specialized suites data docs97.7%
17GPT-5 nano· Stanford HELM specialized suites data docs97.7%
18GPT-4.5 Preview· Stanford HELM specialized suites data docs97.5%
19GPT-5 mini· Stanford HELM specialized suites data docs97.5%
20Granite 3.3 8b Instruct· Stanford HELM specialized suites data docs97.4%
21Palmyra X5· Stanford HELM specialized suites data docs96.5%
22Gemini 2.5 Pro Preview 03-25· Stanford HELM specialized suites data docs96.4%
23Kimi K2 Instruct· Stanford HELM specialized suites data docs95.8%
24Granite 4.0 H Small· Stanford HELM specialized suites data docs95.1%
25gemini-2.5-flash-preview-04-17· Stanford HELM specialized suites data docs94.3%
26GPT 4.1· Stanford HELM specialized suites data docs94.3%
27GLM 4.5 Air· Stanford HELM specialized suites data docs93.9%
28o3 Mini· Stanford HELM specialized suites data docs93.9%
29Gemini 2.0 Flash· Stanford HELM specialized suites data docs93.8%
30GPT-4 Turbo· Stanford HELM specialized suites data docs93.6%
31OpenAI o3· Stanford HELM specialized suites data docs92.3%
32gemini-2.0-pro-exp-02-05· Stanford HELM specialized suites data docs92.2%
33Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs91.9%
34Qwen2 72B Instruct· Stanford HELM specialized suites data docs91.2%
35Gemini 2.0 Flash-Lite Preview· Stanford HELM specialized suites data docs90.7%
36Llama3 8B Chat· Stanford HELM specialized suites data docs90.7%
37o1· Stanford HELM specialized suites data docs90.7%
38gemini-1.5-pro-002· Stanford HELM specialized suites data docs90.6%
39GPT-4.1 mini· Stanford HELM specialized suites data docs90.3%
40o4 Mini· Stanford HELM specialized suites data docs90.1%
41Llama 4 Maverick· Stanford HELM specialized suites data docs90%
42Gemini 2.5 Flash-Lite· Stanford HELM specialized suites data docs89.1%
43Gemini 3 Pro Preview· Stanford HELM specialized suites data docs89.1%
44GPT 4osetting not named by the source· Stanford HELM specialized suites data docs88.7%
45DeepSeek R1· Stanford HELM specialized suites data docs88%
46Llama 3.1 8B Instruct Turbo· Stanford HELM specialized suites data docs87.8%
47Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs87.2%
48gemini-1.5-flash-002· Stanford HELM specialized suites data docs87%
49Grok 3 Mini· Stanford HELM specialized suites data docs84.5%
50GPT-4.1 nano· Stanford HELM specialized suites data docs83.2%

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