Skip to content

Benchmark

HELM MMLU: High School Physics - EM

Scores are percentages. Higher is better.

Models measured
64
Scores recorded
68
Sources
1
IQ weight
1.23

Score distribution

53%
Mean
54.3%
Median
14.3%
Std dev
36.4%
10th percentile
72.8%
90th percentile
25.8%
Min
25.8%83.4%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.23 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 docs83.4%
2DeepSeek V3· Stanford HELM specialized suites data docs80.1%
3gemini-1.5-pro-001· Stanford HELM specialized suites data docs76.2%
4Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs74.2%
5Nova Prov1· Stanford HELM specialized suites data docs74.2%
6GPT 4osetting not named by the source· Stanford HELM specialized suites data docs73.5%
7Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs73.5%
8Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs72.8%
9GPT 4osetting not named by the source· Stanford HELM specialized suites data docs72.8%
10Qwen2 72B Instruct· Stanford HELM specialized suites data docs72.2%
11Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs72.2%
12Claude 3 Opus· Stanford HELM specialized suites data docs70.9%
13gemini-1.5-flash-001· Stanford HELM specialized suites data docs70.2%
14Gemini 2.0 Flash Experimental· Stanford HELM specialized suites data docs68.9%
15Nova Litev1· Stanford HELM specialized suites data docs68.9%
16Llama-3.2-90B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs63.6%
17Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs62.9%
18gemini-1.5-flash-002· Stanford HELM specialized suites data docs61.6%
19Phi-3 Medium 4K Instruct· Stanford HELM specialized suites data docs61.6%
20GPT-4 Turbo· Stanford HELM specialized suites data docs60.3%
21Jamba 1.5 Large· Stanford HELM specialized suites data docs60.3%
22Llama 3.3 70B Instruct Turbo· Stanford HELM specialized suites data docs60.3%
23GPT 4· Stanford HELM specialized suites data docs59.6%
24gemma2-27b· Stanford HELM specialized suites data docs58.9%
25GPT-4o mini· Stanford HELM specialized suites data docs58.9%
26Meta Llama 3 70b· Stanford HELM specialized suites data docs57.6%
27Phi-3 Small 8K Instruct· Stanford HELM specialized suites data docs57.6%
28Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs57.6%
29Qwen1.5-72B· Stanford HELM specialized suites data docs56.3%
30text-unicorn· Stanford HELM specialized suites data docs56.3%
31mixtral-8x22b· Stanford HELM specialized suites data docs55.6%
32Nova Microv1· Stanford HELM specialized suites data docs55%
33Qwen1.5-110B-Chat· Stanford HELM specialized suites data docs55%
34GPT 4 1106 Preview· Stanford HELM specialized suites data docs54.3%
35Mistral Large2407· Stanford HELM specialized suites data docs54.3%
36Claude Haiku 3.5· Stanford HELM specialized suites data docs53%
37Yi-34B· Stanford HELM specialized suites data docs51.7%
38Claude 2.1· Stanford HELM specialized suites data docs47.7%
39Claude 3 Sonnet· Stanford HELM specialized suites data docs47.7%
40DBRX-instruct· Stanford HELM specialized suites data docs47.7%
41Claude 3 Haiku· Stanford HELM specialized suites data docs47%
42gemma-2-9b· Stanford HELM specialized suites data docs46.4%
43Llama 2 70b· Stanford HELM specialized suites data docs46.4%
44Qwen1.5-14B· Stanford HELM specialized suites data docs46.4%
45DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs45.7%
46Claude Instant 1.2· Stanford HELM specialized suites data docs45%
47Mistral Largesetting not named by the source· Stanford HELM specialized suites data docs43.7%
48gemma-7b· Stanford HELM specialized suites data docs41.7%
49Meta Llama 3 8b· Stanford HELM specialized suites data docs41.7%
50Jamba 1.5 Mini· Stanford HELM specialized suites data docs41.1%

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