Skip to content

Benchmark

HELM MMLU: High School Chemistry - EM

Scores are percentages. Higher is better.

Models measured
64
Scores recorded
68
Sources
1
IQ weight
1.60

Score distribution

64.3%
Mean
67%
Median
11.8%
Std dev
49.8%
10th percentile
77.3%
90th percentile
31%
Min
31%83.3%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.60 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
1gemini-1.5-pro-002· Stanford HELM specialized suites data docs83.3%
2DeepSeek V3· Stanford HELM specialized suites data docs82.8%
3Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs82.3%
4Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs81.3%
5Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs80.3%
6GPT 4osetting not named by the source· Stanford HELM specialized suites data docs79.3%
7GPT 4osetting not named by the source· Stanford HELM specialized suites data docs77.8%
8Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs77.3%
9Claude 3 Opus· Stanford HELM specialized suites data docs76.8%
10Nova Prov1· Stanford HELM specialized suites data docs76.8%
11Qwen2 72B Instruct· Stanford HELM specialized suites data docs76.8%
12Llama-3.2-90B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs75.4%
13Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs75.4%
14gemini-1.5-flash-002· Stanford HELM specialized suites data docs74.9%
15text-unicorn· Stanford HELM specialized suites data docs74.9%
16Nova Litev1· Stanford HELM specialized suites data docs74.4%
17Meta Llama 3 70b· Stanford HELM specialized suites data docs73.4%
18Mistral Large2407· Stanford HELM specialized suites data docs73.4%
19Phi-3 Medium 4K Instruct· Stanford HELM specialized suites data docs73.4%
20Gemini 2.0 Flash Experimental· Stanford HELM specialized suites data docs72.9%
21GPT-4o mini· Stanford HELM specialized suites data docs72.9%
22Llama 3.3 70B Instruct Turbo· Stanford HELM specialized suites data docs72.9%
23GPT-4 Turbo· Stanford HELM specialized suites data docs72.4%
24gemini-1.5-pro-001· Stanford HELM specialized suites data docs71.9%
25gemini-1.5-flash-001· Stanford HELM specialized suites data docs71.4%
26GPT 4· Stanford HELM specialized suites data docs70.9%
27Jamba 1.5 Large· Stanford HELM specialized suites data docs70.9%
28gemma2-27b· Stanford HELM specialized suites data docs70.4%
29Claude Haiku 3.5· Stanford HELM specialized suites data docs69%
30mixtral-8x22b· Stanford HELM specialized suites data docs69%
31Phi-3 Small 8K Instruct· Stanford HELM specialized suites data docs68%
32GPT 4 1106 Preview· Stanford HELM specialized suites data docs67.5%
33Qwen1.5-72B· Stanford HELM specialized suites data docs67.5%
34Qwen1.5-110B-Chat· Stanford HELM specialized suites data docs67%
35DBRX-instruct· Stanford HELM specialized suites data docs65.5%
36gemma-2-9b· Stanford HELM specialized suites data docs65.5%
37Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs64.5%
38Yi-34B· Stanford HELM specialized suites data docs63.5%
39Claude 3 Haiku· Stanford HELM specialized suites data docs63.1%
40Qwen1.5-14B· Stanford HELM specialized suites data docs62.6%
41Nova Microv1· Stanford HELM specialized suites data docs61.1%
42Claude 3 Sonnet· Stanford HELM specialized suites data docs60.6%
43Snowflake Arctic Instruct· Stanford HELM specialized suites data docs59.1%
44Claude 2.1· Stanford HELM specialized suites data docs58.6%
45Command R Plus· Stanford HELM specialized suites data docs57.6%
46DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs56.7%
47Jamba 1.5 Mini· Stanford HELM specialized suites data docs56.2%
48GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs55.7%
49GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs55.7%
50Llama 3.1 8B Instruct Turbo· Stanford HELM specialized suites data docs55.2%

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