Skip to content

Benchmark

HELM MMLU: High School Psychology - EM

Scores are percentages. Higher is better.

Models measured
64
Scores recorded
68
Sources
1
IQ weight
2.27

Score distribution

88.9%
Mean
91.7%
Median
9.7%
Std dev
81.5%
10th percentile
95.2%
90th percentile
25.9%
Min
25.9%96.7%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.27 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 docs96.7%
2gemini-1.5-pro-002· Stanford HELM specialized suites data docs96.7%
3Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs96%
4DeepSeek V3· Stanford HELM specialized suites data docs96%
5GPT 4osetting not named by the source· Stanford HELM specialized suites data docs95.6%
6GPT 4 1106 Preview· Stanford HELM specialized suites data docs95.4%
7GPT 4osetting not named by the source· Stanford HELM specialized suites data docs95.4%
8Llama 3.3 70B Instruct Turbo· Stanford HELM specialized suites data docs95.2%
9GPT-4 Turbo· Stanford HELM specialized suites data docs95%
10Nova Prov1· Stanford HELM specialized suites data docs95%
11Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs95%
12gemini-1.5-flash-001· Stanford HELM specialized suites data docs94.9%
13GPT 4· Stanford HELM specialized suites data docs94.9%
14Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs94.9%
15Llama-3.2-90B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs94.9%
16Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs94.9%
17Claude 3 Opus· Stanford HELM specialized suites data docs94.1%
18gemini-1.5-pro-001· Stanford HELM specialized suites data docs94.1%
19Meta Llama 3 70b· Stanford HELM specialized suites data docs93.9%
20Mistral Large2407· Stanford HELM specialized suites data docs93.9%
21text-unicorn· Stanford HELM specialized suites data docs93.8%
22Qwen2 72B Instruct· Stanford HELM specialized suites data docs93.6%
23Nova Litev1· Stanford HELM specialized suites data docs93.4%
24Claude 3 Sonnet· Stanford HELM specialized suites data docs93.2%
25gemma2-27b· Stanford HELM specialized suites data docs93.2%
26Jamba 1.5 Large· Stanford HELM specialized suites data docs93%
27GPT-4o mini· Stanford HELM specialized suites data docs92.8%
28mixtral-8x22b· Stanford HELM specialized suites data docs92.8%
29Qwen1.5-72B· Stanford HELM specialized suites data docs92.8%
30Phi-3 Small 8K Instruct· Stanford HELM specialized suites data docs92.5%
31Phi-3 Medium 4K Instruct· Stanford HELM specialized suites data docs92.3%
32Claude Haiku 3.5· Stanford HELM specialized suites data docs92.1%
33Qwen1.5-110B-Chat· Stanford HELM specialized suites data docs92.1%
34Yi-34B· Stanford HELM specialized suites data docs91.7%
35Claude 3 Haiku· Stanford HELM specialized suites data docs91.6%
36gemma-2-9b· Stanford HELM specialized suites data docs91%
37Mistral Largesetting not named by the source· Stanford HELM specialized suites data docs91%
38DBRX-instruct· Stanford HELM specialized suites data docs90.8%
39DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs90.3%
40Claude 2.1· Stanford HELM specialized suites data docs89.9%
41Nova Microv1· Stanford HELM specialized suites data docs89.9%
42gemini-1.5-flash-002· Stanford HELM specialized suites data docs89.7%
43Mistral Small· Stanford HELM specialized suites data docs88.8%
44Llama 2 70b· Stanford HELM specialized suites data docs88.6%
45Gemini 2.0 Flash Experimental· Stanford HELM specialized suites data docs88.4%
46Jamba 1.5 Mini· Stanford HELM specialized suites data docs88.3%
47Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs87.9%
48Command R Plus· Stanford HELM specialized suites data docs87.3%
49Open Mistral Nemo· Stanford HELM specialized suites data docs87.2%
50GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs87%

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