Skip to content

Benchmark

HELM MMLU: Econometrics - EM

Scores are percentages. Higher is better.

Models measured
64
Scores recorded
68
Sources
1
IQ weight
2.05

Score distribution

59%
Mean
61.4%
Median
12.7%
Std dev
39.5%
10th percentile
72.8%
90th percentile
30.7%
Min
30.7%80.7%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.05 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
1Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs80.7%
2Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs80.7%
3Claude 3 Opus· Stanford HELM specialized suites data docs78.9%
4DeepSeek V3· Stanford HELM specialized suites data docs74.6%
5Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs74.6%
6Qwen2 72B Instruct· Stanford HELM specialized suites data docs73.7%
7gemini-1.5-pro-001· Stanford HELM specialized suites data docs72.8%
8Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs72.8%
9Llama 3.3 70B Instruct Turbo· Stanford HELM specialized suites data docs71.9%
10GPT 4osetting not named by the source· Stanford HELM specialized suites data docs71.1%
11Nova Prov1· Stanford HELM specialized suites data docs70.2%
12Gemini 2.0 Flash Experimental· Stanford HELM specialized suites data docs69.3%
13gemini-1.5-pro-002· Stanford HELM specialized suites data docs69.3%
14GPT 4osetting not named by the source· Stanford HELM specialized suites data docs69.3%
15Meta Llama 3 70b· Stanford HELM specialized suites data docs69.3%
16Mistral Large2407· Stanford HELM specialized suites data docs69.3%
17GPT 4· Stanford HELM specialized suites data docs68.4%
18Llama-3.2-90B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs68.4%
19gemini-1.5-flash-002· Stanford HELM specialized suites data docs67.5%
20GPT 4 1106 Preview· Stanford HELM specialized suites data docs67.5%
21GPT-4 Turbo· Stanford HELM specialized suites data docs67.5%
22Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs67.5%
23Nova Litev1· Stanford HELM specialized suites data docs67.5%
24gemma2-27b· Stanford HELM specialized suites data docs66.7%
25mixtral-8x22b· Stanford HELM specialized suites data docs66.7%
26GPT-4o mini· Stanford HELM specialized suites data docs64.9%
27text-unicorn· Stanford HELM specialized suites data docs64.9%
28Claude 3 Sonnet· Stanford HELM specialized suites data docs64%
29Mistral Largesetting not named by the source· Stanford HELM specialized suites data docs64%
30Qwen1.5-110B-Chat· Stanford HELM specialized suites data docs64%
31Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs64%
32Claude 3 Haiku· Stanford HELM specialized suites data docs63.2%
33Claude Instant 1.2· Stanford HELM specialized suites data docs61.4%
34gemini-1.5-flash-001· Stanford HELM specialized suites data docs61.4%
35Jamba 1.5 Large· Stanford HELM specialized suites data docs61.4%
36Mistral Small· Stanford HELM specialized suites data docs61.4%
37Phi-3 Medium 4K Instruct· Stanford HELM specialized suites data docs61.4%
38DBRX-instruct· Stanford HELM specialized suites data docs60.5%
39Claude 2.1· Stanford HELM specialized suites data docs59.6%
40Claude Haiku 3.5· Stanford HELM specialized suites data docs59.6%
41Phi-3 Small 8K Instruct· Stanford HELM specialized suites data docs59.6%
42Yi-34B· Stanford HELM specialized suites data docs58.8%
43gemma-2-9b· Stanford HELM specialized suites data docs57.9%
44Nova Microv1· Stanford HELM specialized suites data docs57%
45Command R Plus· Stanford HELM specialized suites data docs56.1%
46Open Mistral Nemo· Stanford HELM specialized suites data docs56.1%
47Qwen1.5-14B· Stanford HELM specialized suites data docs56.1%
48DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs55.3%
49Qwen1.5-72B· Stanford HELM specialized suites data docs54.4%
50Meta Llama 3 8b· Stanford HELM specialized suites data docs51.8%

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