Skip to content

Benchmark

HELM AIR-Bench: AIRBench 2024 - #24: Political Persuasion

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
0.69

Score distribution

46.1%
Mean
44.3%
Median
28.6%
Std dev
7.6%
10th percentile
89.5%
90th percentile
4.3%
Min
4.3%93.3%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.69 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
51Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs29%
52GLM 4.5 Air· Stanford HELM specialized suites data docs27.6%
53DeepSeek R1· Stanford HELM specialized suites data docs21.4%
54Yi 34B Chat· Stanford HELM specialized suites data docs21.4%
55GPT 4osetting not named by the source· Stanford HELM specialized suites data docs20.5%
56GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs20.5%
57GPT 4osetting not named by the source· Stanford HELM specialized suites data docs19%
58Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs18.6%
59O1 Mini· Stanford HELM specialized suites data docs16.2%
60DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs15.7%
61GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs14.8%
62Mixtral 8x22B Instruct v0.1· Stanford HELM specialized suites data docs14.8%
63Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs14.3%
64Qwen1.5-72B-Chat· Stanford HELM specialized suites data docs14.3%
65Grok 3· Stanford HELM specialized suites data docs12.4%
66Grok 3 Mini· Stanford HELM specialized suites data docs12.4%
67Mixtral 8x7B Instruct v0.1· Stanford HELM specialized suites data docs11.4%
68DeepSeek V3· Stanford HELM specialized suites data docs7.6%
69Command R+· Stanford HELM specialized suites data docs6.7%
70Command R Plus· Stanford HELM specialized suites data docs6.2%
71Mistral (7B) Instruct v0.3· Stanford HELM specialized suites data docs5.7%
72Mistral Small· Stanford HELM specialized suites data docs5.7%
73Grok 4· Stanford HELM specialized suites data docs4.8%
74Mistral Large2407· Stanford HELM specialized suites data docs4.8%
75DBRX-instruct· Stanford HELM specialized suites data docs4.3%

Showing 51–75 of 75 results.

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