Skip to content

Benchmark

HELM AIR-Bench: AIRBench 2024 - #31: Fraudulent Schemes

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
1.05

Score distribution

70.8%
Mean
76.7%
Median
25.1%
Std dev
33.3%
10th percentile
98.3%
90th percentile
10%
Min
10%100%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.05 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
51DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs63.3%
52GLM 4.5 Air· Stanford HELM specialized suites data docs63.3%
53GPT 4.1· Stanford HELM specialized suites data docs60%
54GPT 4osetting not named by the source· Stanford HELM specialized suites data docs60%
55Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs60%
56Qwen1.5-72B-Chat· Stanford HELM specialized suites data docs60%
57Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs60%
58Llama 4 Scout· Stanford HELM specialized suites data docs51.7%
59O1 Mini· Stanford HELM specialized suites data docs50%
60GPT 4osetting not named by the source· Stanford HELM specialized suites data docs46.7%
61Grok 3 Mini· Stanford HELM specialized suites data docs46.7%
62GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs43.3%
63Grok 3· Stanford HELM specialized suites data docs40%
64GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs36.7%
65Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs36.7%
66Mixtral 8x7B Instruct v0.1· Stanford HELM specialized suites data docs35%
67DeepSeek R1· Stanford HELM specialized suites data docs33.3%
68Mistral Small· Stanford HELM specialized suites data docs33.3%
69DeepSeek V3· Stanford HELM specialized suites data docs28.3%
70Command R+· Stanford HELM specialized suites data docs23.3%
71Grok 4· Stanford HELM specialized suites data docs23.3%
72Mistral Large2407· Stanford HELM specialized suites data docs20%
73Command R Plus· Stanford HELM specialized suites data docs13.3%
74Mistral (7B) Instruct v0.3· Stanford HELM specialized suites data docs13.3%
75DBRX-instruct· Stanford HELM specialized suites data docs10%

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