Skip to content

Benchmark

HELM MMLU: Nutrition - EM

Scores are percentages. Higher is better.

Models measured
64
Scores recorded
68
Sources
1
IQ weight
1.89

Score distribution

79.6%
Mean
82%
Median
10.4%
Std dev
69.6%
10th percentile
89.9%
90th percentile
34%
Min
34%92.8%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.89 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
1Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs92.8%
2Claude 3 Opus· Stanford HELM specialized suites data docs92.5%
3Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs92.2%
4DeepSeek V3· Stanford HELM specialized suites data docs91.8%
5Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs91.2%
6GPT 4osetting not named by the source· Stanford HELM specialized suites data docs90.5%
7Qwen2 72B Instruct· Stanford HELM specialized suites data docs90.2%
8GPT 4osetting not named by the source· Stanford HELM specialized suites data docs89.9%
9GPT 4· Stanford HELM specialized suites data docs89.2%
10GPT-4 Turbo· Stanford HELM specialized suites data docs89.2%
11Llama-3.2-90B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs88.9%
12Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs88.9%
13gemini-1.5-pro-002· Stanford HELM specialized suites data docs88.6%
14Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs88.6%
15Llama 3.3 70B Instruct Turbo· Stanford HELM specialized suites data docs88.2%
16gemini-1.5-pro-001· Stanford HELM specialized suites data docs87.9%
17GPT 4 1106 Preview· Stanford HELM specialized suites data docs87.9%
18Meta Llama 3 70b· Stanford HELM specialized suites data docs87.6%
19Jamba 1.5 Large· Stanford HELM specialized suites data docs86.9%
20Yi-34B· Stanford HELM specialized suites data docs86.9%
21mixtral-8x22b· Stanford HELM specialized suites data docs86.6%
22Nova Prov1· Stanford HELM specialized suites data docs86.6%
23Qwen1.5-72B· Stanford HELM specialized suites data docs85.9%
24Gemini 2.0 Flash Experimental· Stanford HELM specialized suites data docs85.6%
25text-unicorn· Stanford HELM specialized suites data docs85.6%
26Claude Haiku 3.5· Stanford HELM specialized suites data docs84.6%
27Phi-3 Medium 4K Instruct· Stanford HELM specialized suites data docs83.7%
28Phi-3 Small 8K Instruct· Stanford HELM specialized suites data docs83.3%
29Claude 3 Haiku· Stanford HELM specialized suites data docs83%
30GPT-4o mini· Stanford HELM specialized suites data docs82.7%
31Mistral Large2407· Stanford HELM specialized suites data docs82.7%
32gemma2-27b· Stanford HELM specialized suites data docs82.4%
33Claude 3 Sonnet· Stanford HELM specialized suites data docs82%
34gemini-1.5-flash-001· Stanford HELM specialized suites data docs82%
35DBRX-instruct· Stanford HELM specialized suites data docs81.4%
36Qwen1.5-110B-Chat· Stanford HELM specialized suites data docs80.4%
37Jamba 1.5 Mini· Stanford HELM specialized suites data docs80.1%
38Mistral Largesetting not named by the source· Stanford HELM specialized suites data docs79.1%
39Nova Litev1· Stanford HELM specialized suites data docs78.8%
40Claude 2.1· Stanford HELM specialized suites data docs78.1%
41DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs78.1%
42gemma-7b· Stanford HELM specialized suites data docs77.8%
43Nova Microv1· Stanford HELM specialized suites data docs77.8%
44Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs77.8%
45gemma-2-9b· Stanford HELM specialized suites data docs77.5%
46Meta Llama 3 8b· Stanford HELM specialized suites data docs76.1%
47Mistral Small· Stanford HELM specialized suites data docs76.1%
48GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs75.8%
49Llama 2 70b· Stanford HELM specialized suites data docs75.8%
50Llama-3.2-11B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs75.2%

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