Skip to content

Benchmark

HELM MMLU: Business Ethics - EM

Scores are percentages. Higher is better.

Models measured
64
Scores recorded
68
Sources
1
IQ weight
2.16

Score distribution

73.2%
Mean
77%
Median
12.3%
Std dev
59%
10th percentile
84%
90th percentile
24%
Min
24%89%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.16 for chat models, measured over 64 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1DeepSeek V3· Stanford HELM specialized suites data docs89%
2GPT 4osetting not named by the source· Stanford HELM specialized suites data docs89%
3Claude 3 Opus· Stanford HELM specialized suites data docs86%
4DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs86%
5Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs85%
6GPT 4osetting not named by the source· Stanford HELM specialized suites data docs85%
7Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs85%
8gemini-1.5-pro-002· Stanford HELM specialized suites data docs84%
9Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs83%
10Meta Llama 3 70b· Stanford HELM specialized suites data docs83%
11text-unicorn· Stanford HELM specialized suites data docs83%
12Claude 3 Sonnet· Stanford HELM specialized suites data docs82%
13GPT-4 Turbo· Stanford HELM specialized suites data docs82%
14Qwen2 72B Instruct· Stanford HELM specialized suites data docs82%
15Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs82%
16gemini-1.5-flash-001· Stanford HELM specialized suites data docs81%
17Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs81%
18Nova Prov1· Stanford HELM specialized suites data docs81%
19Claude Haiku 3.5· Stanford HELM specialized suites data docs80%
20gemini-1.5-pro-001· Stanford HELM specialized suites data docs80%
21Llama 3.3 70B Instruct Turbo· Stanford HELM specialized suites data docs80%
22Phi-3 Medium 4K Instruct· Stanford HELM specialized suites data docs80%
23Qwen1.5-110B-Chat· Stanford HELM specialized suites data docs80%
24GPT 4· Stanford HELM specialized suites data docs79%
25GPT-4o mini· Stanford HELM specialized suites data docs79%
26Mistral Large2407· Stanford HELM specialized suites data docs79%
27Qwen1.5-72B· Stanford HELM specialized suites data docs79%
28Claude 3 Haiku· Stanford HELM specialized suites data docs78%
29DBRX-instruct· Stanford HELM specialized suites data docs78%
30gemma2-27b· Stanford HELM specialized suites data docs78%
31GPT 4 1106 Preview· Stanford HELM specialized suites data docs78%
32Command R Plus· Stanford HELM specialized suites data docs77%
33gemma-2-9b· Stanford HELM specialized suites data docs77%
34Jamba 1.5 Large· Stanford HELM specialized suites data docs77%
35Phi-3 Small 8K Instruct· Stanford HELM specialized suites data docs77%
36Jamba 1.5 Mini· Stanford HELM specialized suites data docs76%
37Llama-3.2-90B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs76%
38GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs75%
39GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs75%
40Qwen1.5-14B· Stanford HELM specialized suites data docs75%
41Yi-34B· Stanford HELM specialized suites data docs75%
42mixtral-8x22b· Stanford HELM specialized suites data docs74%
43Claude 2.1· Stanford HELM specialized suites data docs73%
44Gemini 2.0 Flash Experimental· Stanford HELM specialized suites data docs73%
45Llama 2 70b· Stanford HELM specialized suites data docs73%
46Nova Litev1· Stanford HELM specialized suites data docs73%
47Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs72%
48Mistral Small· Stanford HELM specialized suites data docs71%
49Nova Microv1· Stanford HELM specialized suites data docs71%
50Claude Instant 1.2· Stanford HELM specialized suites data docs70%

Showing 1–50 of 68 results.

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