Skip to content

Benchmark

HELM MMLU: MMLU All Subjects - EM

Scores are percentages. Higher is better.

Models measured
64
Scores recorded
68
Sources
1
IQ weight
2.11

Score distribution

72.9%
Mean
75.7%
Median
10.3%
Std dev
59.9%
10th percentile
84.2%
90th percentile
29.5%
Min
29.5%87.3%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.11 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 docs87.3%
2DeepSeek V3· Stanford HELM specialized suites data docs87.2%
3gemini-1.5-pro-002· Stanford HELM specialized suites data docs86.9%
4Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs86.5%
5Claude 3 Opus· Stanford HELM specialized suites data docs84.6%
6Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs84.5%
7GPT 4osetting not named by the source· Stanford HELM specialized suites data docs84.3%
8GPT 4osetting not named by the source· Stanford HELM specialized suites data docs84.2%
9Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs83.4%
10gemini-1.5-pro-001· Stanford HELM specialized suites data docs82.7%
11GPT 4· Stanford HELM specialized suites data docs82.4%
12Qwen2 72B Instruct· Stanford HELM specialized suites data docs82.4%
13Nova Prov1· Stanford HELM specialized suites data docs82%
14GPT-4 Turbo· Stanford HELM specialized suites data docs81.3%
15Llama-3.2-90B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs80.3%
16Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs80.1%
17Mistral Large2407· Stanford HELM specialized suites data docs80%
18Gemini 2.0 Flash Experimental· Stanford HELM specialized suites data docs79.7%
19GPT 4 1106 Preview· Stanford HELM specialized suites data docs79.6%
20Meta Llama 3 70b· Stanford HELM specialized suites data docs79.3%
21Llama 3.3 70B Instruct Turbo· Stanford HELM specialized suites data docs79.1%
22text-unicorn· Stanford HELM specialized suites data docs78.6%
23Jamba 1.5 Large· Stanford HELM specialized suites data docs78.2%
24gemini-1.5-flash-001· Stanford HELM specialized suites data docs77.9%
25mixtral-8x22b· Stanford HELM specialized suites data docs77.8%
26Phi-3 Medium 4K Instruct· Stanford HELM specialized suites data docs77.5%
27Qwen1.5-72B· Stanford HELM specialized suites data docs77.4%
28Nova Litev1· Stanford HELM specialized suites data docs77%
29Qwen1.5-110B-Chat· Stanford HELM specialized suites data docs76.8%
30GPT-4o mini· Stanford HELM specialized suites data docs76.7%
31Yi-34B· Stanford HELM specialized suites data docs76.2%
32Claude 3 Sonnet· Stanford HELM specialized suites data docs75.9%
33gemma2-27b· Stanford HELM specialized suites data docs75.7%
34Phi-3 Small 8K Instruct· Stanford HELM specialized suites data docs75.7%
35Claude Haiku 3.5· Stanford HELM specialized suites data docs74.3%
36DBRX-instruct· Stanford HELM specialized suites data docs74.1%
37gemini-1.5-flash-002· Stanford HELM specialized suites data docs73.9%
38Claude 3 Haiku· Stanford HELM specialized suites data docs73.8%
39Claude 2.1· Stanford HELM specialized suites data docs73.5%
40Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs72.9%
41DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs72.5%
42gemma-2-9b· Stanford HELM specialized suites data docs72.1%
43Nova Microv1· Stanford HELM specialized suites data docs70.8%
44Jamba 1.5 Mini· Stanford HELM specialized suites data docs69.9%
45Llama 2 70b· Stanford HELM specialized suites data docs69.5%
46Command R Plus· Stanford HELM specialized suites data docs69.4%
47GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs68.9%
48Claude Instant 1.2· Stanford HELM specialized suites data docs68.8%
49Mistral Largesetting not named by the source· Stanford HELM specialized suites data docs68.8%
50Mistral Small· Stanford HELM specialized suites data docs68.7%

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