Skip to content

Benchmark

HELM MMLU: Human Aging - EM

Scores are percentages. Higher is better.

Models measured
64
Scores recorded
68
Sources
1
IQ weight
2.16

Score distribution

75.9%
Mean
78%
Median
9.4%
Std dev
64.1%
10th percentile
84.3%
90th percentile
27.8%
Min
27.8%88.8%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.16 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 docs88.8%
2Claude 3 Opus· Stanford HELM specialized suites data docs86.1%
3Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs86.1%
4DeepSeek V3· Stanford HELM specialized suites data docs86.1%
5GPT 4 1106 Preview· Stanford HELM specialized suites data docs85.7%
6GPT 4osetting not named by the source· Stanford HELM specialized suites data docs84.8%
7GPT-4 Turbo· Stanford HELM specialized suites data docs84.8%
8GPT 4osetting not named by the source· Stanford HELM specialized suites data docs84.3%
9Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs84.3%
10gemini-1.5-pro-002· Stanford HELM specialized suites data docs83.9%
11GPT 4· Stanford HELM specialized suites data docs83.4%
12Llama-3.2-90B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs83.4%
13Nova Prov1· Stanford HELM specialized suites data docs83%
14DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs82.5%
15Meta Llama 3 70b· Stanford HELM specialized suites data docs82.5%
16gemma2-27b· Stanford HELM specialized suites data docs82.1%
17Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs82.1%
18Qwen2 72B Instruct· Stanford HELM specialized suites data docs81.6%
19Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs81.6%
20GPT-4o mini· Stanford HELM specialized suites data docs81.2%
21Jamba 1.5 Large· Stanford HELM specialized suites data docs81.2%
22Llama 3.3 70B Instruct Turbo· Stanford HELM specialized suites data docs81.2%
23Qwen1.5-72B· Stanford HELM specialized suites data docs81.2%
24gemini-1.5-pro-001· Stanford HELM specialized suites data docs80.7%
25Mistral Large2407· Stanford HELM specialized suites data docs80.7%
26Mistral Largesetting not named by the source· Stanford HELM specialized suites data docs80.3%
27gemini-1.5-flash-001· Stanford HELM specialized suites data docs79.8%
28Qwen1.5-110B-Chat· Stanford HELM specialized suites data docs79.8%
29Claude 3 Sonnet· Stanford HELM specialized suites data docs79.4%
30Llama 2 70b· Stanford HELM specialized suites data docs79.4%
31text-unicorn· Stanford HELM specialized suites data docs79.4%
32Yi-34B· Stanford HELM specialized suites data docs78.9%
33Phi-3 Small 8K Instruct· Stanford HELM specialized suites data docs78.5%
34mixtral-8x22b· Stanford HELM specialized suites data docs78%
35Nova Litev1· Stanford HELM specialized suites data docs78%
36Phi-3 Medium 4K Instruct· Stanford HELM specialized suites data docs78%
37Claude 2.1· Stanford HELM specialized suites data docs77.6%
38Claude 3 Haiku· Stanford HELM specialized suites data docs76.7%
39DBRX-instruct· Stanford HELM specialized suites data docs76.7%
40Claude Haiku 3.5· Stanford HELM specialized suites data docs76.2%
41GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs76.2%
42gemma-2-9b· Stanford HELM specialized suites data docs75.3%
43GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs74.9%
44Jamba 1.5 Mini· Stanford HELM specialized suites data docs74.9%
45Mistral Small· Stanford HELM specialized suites data docs74.4%
46Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs74.4%
47Command R Plus· Stanford HELM specialized suites data docs74%
48gemini-1.5-flash-002· Stanford HELM specialized suites data docs74%
49Qwen1.5-14B· Stanford HELM specialized suites data docs74%
50Command R+· Stanford HELM specialized suites data docs73.1%

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