Skip to content

Benchmark

HELM MMLU: College Mathematics - EM

Scores are percentages. Higher is better.

Models measured
64
Scores recorded
68
Sources
1
IQ weight
1.51

Score distribution

45.7%
Mean
46%
Median
12.4%
Std dev
30%
10th percentile
60%
90th percentile
22%
Min
22%80%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.51 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 docs80%
2gemini-1.5-pro-001· Stanford HELM specialized suites data docs79%
3DeepSeek V3· Stanford HELM specialized suites data docs76%
4Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs66%
5Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs66%
6Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs64%
7Gemini 2.0 Flash Experimental· Stanford HELM specialized suites data docs61%
8Nova Prov1· Stanford HELM specialized suites data docs60%
9Mistral Large2407· Stanford HELM specialized suites data docs59%
10Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs58%
11Meta Llama 3 70b· Stanford HELM specialized suites data docs56%
12Qwen1.5-72B· Stanford HELM specialized suites data docs56%
13Claude 3 Opus· Stanford HELM specialized suites data docs55%
14GPT-4 Turbo· Stanford HELM specialized suites data docs55%
15Qwen1.5-110B-Chat· Stanford HELM specialized suites data docs55%
16gemini-1.5-flash-001· Stanford HELM specialized suites data docs54%
17GPT 4· Stanford HELM specialized suites data docs54%
18Nova Litev1· Stanford HELM specialized suites data docs54%
19Phi-3 Medium 4K Instruct· Stanford HELM specialized suites data docs53%
20gemini-1.5-flash-002· Stanford HELM specialized suites data docs52%
21Qwen2 72B Instruct· Stanford HELM specialized suites data docs52%
22GPT 4osetting not named by the source· Stanford HELM specialized suites data docs51%
23text-unicorn· Stanford HELM specialized suites data docs51%
24Llama-3.2-90B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs50%
25Claude 3 Haiku· Stanford HELM specialized suites data docs49%
26Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs49%
27mixtral-8x22b· Stanford HELM specialized suites data docs48%
28Phi-3 Small 8K Instruct· Stanford HELM specialized suites data docs48%
29GPT 4osetting not named by the source· Stanford HELM specialized suites data docs47%
30GPT-4o mini· Stanford HELM specialized suites data docs47%
31Yi-34B· Stanford HELM specialized suites data docs47%
32Jamba 1.5 Large· Stanford HELM specialized suites data docs46%
33Jamba 1.5 Mini· Stanford HELM specialized suites data docs46%
34Llama 3.3 70B Instruct Turbo· Stanford HELM specialized suites data docs46%
35Nova Microv1· Stanford HELM specialized suites data docs44%
36Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs44%
37gemma2-27b· Stanford HELM specialized suites data docs43%
38DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs42%
39gemma-2-9b· Stanford HELM specialized suites data docs42%
40Qwen1.5-14B· Stanford HELM specialized suites data docs42%
41Claude 3 Sonnet· Stanford HELM specialized suites data docs41%
42Claude Instant 1.2· Stanford HELM specialized suites data docs40%
43GPT 4 1106 Preview· Stanford HELM specialized suites data docs40%
44phi-2· Stanford HELM specialized suites data docs40%
45DBRX-instruct· Stanford HELM specialized suites data docs39%
46Jamba Instruct· Stanford HELM specialized suites data docs39%
47Llama 2 70b· Stanford HELM specialized suites data docs39%
48Open Mistral Nemo· Stanford HELM specialized suites data docs39%
49Claude 2.1· Stanford HELM specialized suites data docs37%
50Command R Plus· Stanford HELM specialized suites data docs37%

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