Skip to content

Benchmark

HELM MMLU: Elementary Mathematics - EM

Scores are percentages. Higher is better.

Models measured
64
Scores recorded
68
Sources
1
IQ weight
1.49

Score distribution

61.5%
Mean
62.2%
Median
17%
Std dev
42.6%
10th percentile
85.7%
90th percentile
25.4%
Min
25.4%94.2%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.49 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
1DeepSeek V3· Stanford HELM specialized suites data docs94.2%
2gemini-1.5-pro-002· Stanford HELM specialized suites data docs94.2%
3gemini-1.5-pro-001· Stanford HELM specialized suites data docs93.9%
4Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs91.8%
5Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs89.2%
6Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs87%
7Claude 3 Opus· Stanford HELM specialized suites data docs86.2%
8Gemini 2.0 Flash Experimental· Stanford HELM specialized suites data docs85.7%
9Nova Prov1· Stanford HELM specialized suites data docs83.1%
10Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs82.8%
11Qwen2 72B Instruct· Stanford HELM specialized suites data docs82.5%
12GPT 4· Stanford HELM specialized suites data docs80.7%
13Mistral Large2407· Stanford HELM specialized suites data docs79.9%
14GPT 4osetting not named by the source· Stanford HELM specialized suites data docs77.5%
15Nova Litev1· Stanford HELM specialized suites data docs75.7%
16gemini-1.5-flash-001· Stanford HELM specialized suites data docs75.4%
17GPT 4osetting not named by the source· Stanford HELM specialized suites data docs74.1%
18GPT-4 Turbo· Stanford HELM specialized suites data docs72%
19Phi-3 Medium 4K Instruct· Stanford HELM specialized suites data docs70.9%
20gemini-1.5-flash-002· Stanford HELM specialized suites data docs70.4%
21Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs70.1%
22Qwen1.5-72B· Stanford HELM specialized suites data docs69.6%
23Llama-3.2-90B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs68.8%
24Llama 3.3 70B Instruct Turbo· Stanford HELM specialized suites data docs67.2%
25Qwen1.5-110B-Chat· Stanford HELM specialized suites data docs66.9%
26text-unicorn· Stanford HELM specialized suites data docs66.1%
27Jamba 1.5 Large· Stanford HELM specialized suites data docs65.6%
28Yi-34B· Stanford HELM specialized suites data docs65.6%
29GPT-4o mini· Stanford HELM specialized suites data docs65.1%
30Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs64.3%
31GPT 4 1106 Preview· Stanford HELM specialized suites data docs63.8%
32Claude 3 Sonnet· Stanford HELM specialized suites data docs63.5%
33Meta Llama 3 70b· Stanford HELM specialized suites data docs63.2%
34mixtral-8x22b· Stanford HELM specialized suites data docs62.2%
35Phi-3 Small 8K Instruct· Stanford HELM specialized suites data docs61.9%
36Qwen1.5-14B· Stanford HELM specialized suites data docs60.3%
37gemma-2-9b· Stanford HELM specialized suites data docs57.7%
38DBRX-instruct· Stanford HELM specialized suites data docs56.3%
39Claude Haiku 3.5· Stanford HELM specialized suites data docs56.1%
40Claude 3 Haiku· Stanford HELM specialized suites data docs55.8%
41gemma2-27b· Stanford HELM specialized suites data docs55.8%
42Jamba 1.5 Mini· Stanford HELM specialized suites data docs55.3%
43Nova Microv1· Stanford HELM specialized suites data docs55%
44DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs54.8%
45GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs53.4%
46Claude 2.1· Stanford HELM specialized suites data docs52.1%
47gemma-7b· Stanford HELM specialized suites data docs51.6%
48Mistral Largesetting not named by the source· Stanford HELM specialized suites data docs50.8%
49GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs50%
50Qwen1.5-7B· Stanford HELM specialized suites data docs50%

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