Skip to content

Benchmark

HELM MMLU: High School Government And Politics - EM

Scores are percentages. Higher is better.

Models measured
64
Scores recorded
68
Sources
1
IQ weight
2.44

Score distribution

93%
Mean
96.4%
Median
9%
Std dev
86.5%
10th percentile
98.4%
90th percentile
35.8%
Min
35.8%99%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.44 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 docs99%
2GPT 4· Stanford HELM specialized suites data docs99%
3Qwen1.5-72B· Stanford HELM specialized suites data docs99%
4Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs99%
5Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs98.4%
6GPT 4osetting not named by the source· Stanford HELM specialized suites data docs98.4%
7Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs98.4%
8Llama 3.3 70B Instruct Turbo· Stanford HELM specialized suites data docs98.4%
9Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs98.4%
10Qwen1.5-110B-Chat· Stanford HELM specialized suites data docs98.4%
11Claude 3 Opus· Stanford HELM specialized suites data docs97.9%
12Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs97.9%
13DeepSeek V3· Stanford HELM specialized suites data docs97.9%
14gemma2-27b· Stanford HELM specialized suites data docs97.9%
15GPT 4 1106 Preview· Stanford HELM specialized suites data docs97.9%
16GPT 4osetting not named by the source· Stanford HELM specialized suites data docs97.9%
17Llama-3.2-90B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs97.9%
18Meta Llama 3 70b· Stanford HELM specialized suites data docs97.9%
19Qwen2 72B Instruct· Stanford HELM specialized suites data docs97.9%
20text-unicorn· Stanford HELM specialized suites data docs97.9%
21gemini-1.5-flash-001· Stanford HELM specialized suites data docs97.4%
22gemini-1.5-pro-001· Stanford HELM specialized suites data docs97.4%
23GPT-4 Turbo· Stanford HELM specialized suites data docs97.4%
24mixtral-8x22b· Stanford HELM specialized suites data docs97.4%
25Nova Prov1· Stanford HELM specialized suites data docs97.4%
26Yi-34B· Stanford HELM specialized suites data docs97.4%
27Gemini 2.0 Flash Experimental· Stanford HELM specialized suites data docs96.9%
28Jamba 1.5 Large· Stanford HELM specialized suites data docs96.9%
29Mistral Large2407· Stanford HELM specialized suites data docs96.9%
30Nova Litev1· Stanford HELM specialized suites data docs96.9%
31Phi-3 Medium 4K Instruct· Stanford HELM specialized suites data docs96.9%
32Phi-3 Small 8K Instruct· Stanford HELM specialized suites data docs96.9%
33DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs96.4%
34Mistral Largesetting not named by the source· Stanford HELM specialized suites data docs96.4%
35Claude 2.1· Stanford HELM specialized suites data docs95.9%
36Claude 3 Sonnet· Stanford HELM specialized suites data docs95.9%
37gemini-1.5-flash-002· Stanford HELM specialized suites data docs95.9%
38GPT-4o mini· Stanford HELM specialized suites data docs95.9%
39DBRX-instruct· Stanford HELM specialized suites data docs95.3%
40gemma-2-9b· Stanford HELM specialized suites data docs95.3%
41Claude 3 Haiku· Stanford HELM specialized suites data docs94.8%
42Mistral Small· Stanford HELM specialized suites data docs94.8%
43Jamba 1.5 Mini· Stanford HELM specialized suites data docs94.3%
44Llama 2 70b· Stanford HELM specialized suites data docs93.3%
45Command R Plus· Stanford HELM specialized suites data docs92.2%
46GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs92.2%
47GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs92.2%
48Nova Microv1· Stanford HELM specialized suites data docs92.2%
49Claude Haiku 3.5· Stanford HELM specialized suites data docs91.7%
50Open Mistral Nemo· Stanford HELM specialized suites data docs91.2%

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