Skip to content

Benchmark

HELM MMLU: Moral Disputes - EM

Scores are percentages. Higher is better.

Models measured
64
Scores recorded
68
Sources
1
IQ weight
1.98

Score distribution

77.4%
Mean
80.6%
Median
10.3%
Std dev
66.8%
10th percentile
86.7%
90th percentile
27.2%
Min
27.2%88.7%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.98 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
1Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs88.7%
2Claude 3 Opus· Stanford HELM specialized suites data docs88.2%
3GPT 4osetting not named by the source· Stanford HELM specialized suites data docs88.2%
4DeepSeek V3· Stanford HELM specialized suites data docs87.3%
5Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs87%
6GPT 4osetting not named by the source· Stanford HELM specialized suites data docs87%
7Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs87%
8GPT 4· Stanford HELM specialized suites data docs86.7%
9Gemini 2.0 Flash Experimental· Stanford HELM specialized suites data docs86.4%
10GPT-4 Turbo· Stanford HELM specialized suites data docs86.1%
11GPT 4 1106 Preview· Stanford HELM specialized suites data docs85.8%
12gemini-1.5-pro-002· Stanford HELM specialized suites data docs85.5%
13Llama 3.3 70B Instruct Turbo· Stanford HELM specialized suites data docs85.3%
14Qwen2 72B Instruct· Stanford HELM specialized suites data docs85.3%
15Jamba 1.5 Large· Stanford HELM specialized suites data docs85%
16Llama-3.2-90B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs85%
17Meta Llama 3 70b· Stanford HELM specialized suites data docs85%
18Nova Prov1· Stanford HELM specialized suites data docs84.7%
19Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs84.1%
20gemma2-27b· Stanford HELM specialized suites data docs83.8%
21Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs83.8%
22Mistral Large2407· Stanford HELM specialized suites data docs83.2%
23mixtral-8x22b· Stanford HELM specialized suites data docs83.2%
24Qwen1.5-72B· Stanford HELM specialized suites data docs83.2%
25text-unicorn· Stanford HELM specialized suites data docs83.2%
26Claude 3 Sonnet· Stanford HELM specialized suites data docs82.9%
27Mistral Largesetting not named by the source· Stanford HELM specialized suites data docs82.9%
28Yi-34B· Stanford HELM specialized suites data docs82.7%
29gemini-1.5-pro-001· Stanford HELM specialized suites data docs81.8%
30DBRX-instruct· Stanford HELM specialized suites data docs81.5%
31gemini-1.5-flash-001· Stanford HELM specialized suites data docs81.2%
32Claude 2.1· Stanford HELM specialized suites data docs80.9%
33Claude 3 Haiku· Stanford HELM specialized suites data docs80.9%
34GPT-4o mini· Stanford HELM specialized suites data docs80.6%
35Claude Haiku 3.5· Stanford HELM specialized suites data docs79.8%
36DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs79.8%
37Phi-3 Small 8K Instruct· Stanford HELM specialized suites data docs79.5%
38Nova Litev1· Stanford HELM specialized suites data docs79.2%
39Jamba 1.5 Mini· Stanford HELM specialized suites data docs78.3%
40gemma-2-9b· Stanford HELM specialized suites data docs77.7%
41Qwen1.5-110B-Chat· Stanford HELM specialized suites data docs77.7%
42Llama 2 70b· Stanford HELM specialized suites data docs77.5%
43Phi-3 Medium 4K Instruct· Stanford HELM specialized suites data docs76.9%
44Claude Instant 1.2· Stanford HELM specialized suites data docs75.7%
45GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs75.7%
46Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs75.7%
47Snowflake Arctic Instruct· Stanford HELM specialized suites data docs74.9%
48Mistral Small· Stanford HELM specialized suites data docs74.3%
49GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs73.4%
50Yi-6B· Stanford HELM specialized suites data docs73.1%

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