Skip to content

Benchmark

HELM MMLU: Conceptual Physics - EM

Scores are percentages. Higher is better.

Models measured
64
Scores recorded
68
Sources
1
IQ weight
1.84

Score distribution

73.5%
Mean
77.4%
Median
14.4%
Std dev
53.6%
10th percentile
89.4%
90th percentile
31.9%
Min
31.9%94.9%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.84 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-001· Stanford HELM specialized suites data docs94.9%
2gemini-1.5-pro-002· Stanford HELM specialized suites data docs94.5%
3DeepSeek V3· Stanford HELM specialized suites data docs94%
4GPT 4osetting not named by the source· Stanford HELM specialized suites data docs92.3%
5GPT 4osetting not named by the source· Stanford HELM specialized suites data docs91.1%
6Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs90.6%
7GPT 4 1106 Preview· Stanford HELM specialized suites data docs89.4%
8GPT-4 Turbo· Stanford HELM specialized suites data docs89.4%
9Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs88.5%
10Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs88.5%
11Claude 3 Opus· Stanford HELM specialized suites data docs88.1%
12Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs87.7%
13Qwen2 72B Instruct· Stanford HELM specialized suites data docs87.2%
14GPT 4· Stanford HELM specialized suites data docs86.8%
15Mistral Large2407· Stanford HELM specialized suites data docs86.4%
16gemini-1.5-flash-001· Stanford HELM specialized suites data docs85.1%
17gemini-1.5-flash-002· Stanford HELM specialized suites data docs85.1%
18Nova Prov1· Stanford HELM specialized suites data docs85.1%
19Meta Llama 3 70b· Stanford HELM specialized suites data docs83.8%
20Qwen1.5-110B-Chat· Stanford HELM specialized suites data docs83.8%
21gemma2-27b· Stanford HELM specialized suites data docs83.4%
22Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs83.4%
23Llama-3.2-90B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs82.6%
24Llama 3.3 70B Instruct Turbo· Stanford HELM specialized suites data docs82.1%
25Qwen1.5-72B· Stanford HELM specialized suites data docs82.1%
26Gemini 2.0 Flash Experimental· Stanford HELM specialized suites data docs81.3%
27Phi-3 Medium 4K Instruct· Stanford HELM specialized suites data docs80.9%
28text-unicorn· Stanford HELM specialized suites data docs80.9%
29mixtral-8x22b· Stanford HELM specialized suites data docs79.6%
30Nova Litev1· Stanford HELM specialized suites data docs79.6%
31GPT-4o mini· Stanford HELM specialized suites data docs79.1%
32Jamba 1.5 Large· Stanford HELM specialized suites data docs77.9%
33Phi-3 Small 8K Instruct· Stanford HELM specialized suites data docs77.9%
34Claude 3 Sonnet· Stanford HELM specialized suites data docs77.4%
35Yi-34B· Stanford HELM specialized suites data docs77%
36Claude 2.1· Stanford HELM specialized suites data docs76.6%
37DBRX-instruct· Stanford HELM specialized suites data docs74%
38Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs73.6%
39gemma-2-9b· Stanford HELM specialized suites data docs73.2%
40Claude Haiku 3.5· Stanford HELM specialized suites data docs72.3%
41DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs72.3%
42Claude 3 Haiku· Stanford HELM specialized suites data docs71.5%
43Nova Microv1· Stanford HELM specialized suites data docs70.6%
44Qwen1.5-14B· Stanford HELM specialized suites data docs69.4%
45Mistral Small· Stanford HELM specialized suites data docs68.5%
46Jamba 1.5 Mini· Stanford HELM specialized suites data docs67.7%
47Jamba Instruct· Stanford HELM specialized suites data docs67.7%
48Llama 2 70b· Stanford HELM specialized suites data docs66.8%
49Open Mistral Nemo· Stanford HELM specialized suites data docs64.7%
50GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs63.4%

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