Skip to content

Benchmark

HELM MMLU: High School World History - EM

Scores are percentages. Higher is better.

Models measured
64
Scores recorded
68
Sources
1
IQ weight
2.24

Score distribution

85.5%
Mean
89%
Median
11.6%
Std dev
74.3%
10th percentile
94.1%
90th percentile
25.3%
Min
25.3%95.8%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.24 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 docs95.8%
2GPT 4 1106 Preview· Stanford HELM specialized suites data docs95.8%
3Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs95.4%
4GPT 4· Stanford HELM specialized suites data docs94.5%
5GPT 4osetting not named by the source· Stanford HELM specialized suites data docs94.5%
6Claude 3 Opus· Stanford HELM specialized suites data docs94.1%
7GPT 4osetting not named by the source· Stanford HELM specialized suites data docs94.1%
8GPT-4 Turbo· Stanford HELM specialized suites data docs94.1%
9Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs94.1%
10Llama-3.2-90B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs94.1%
11Meta Llama 3 70b· Stanford HELM specialized suites data docs94.1%
12gemini-1.5-pro-002· Stanford HELM specialized suites data docs93.7%
13Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs93.7%
14Qwen2 72B Instruct· Stanford HELM specialized suites data docs93.2%
15DeepSeek V3· Stanford HELM specialized suites data docs92.8%
16Nova Prov1· Stanford HELM specialized suites data docs92.8%
17gemini-1.5-pro-001· Stanford HELM specialized suites data docs92.4%
18Mistral Large2407· Stanford HELM specialized suites data docs92%
19Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs92%
20DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs91.1%
21Jamba 1.5 Large· Stanford HELM specialized suites data docs91.1%
22text-unicorn· Stanford HELM specialized suites data docs91.1%
23gemini-1.5-flash-001· Stanford HELM specialized suites data docs90.7%
24Llama 3.3 70B Instruct Turbo· Stanford HELM specialized suites data docs90.7%
25Yi-34B· Stanford HELM specialized suites data docs90.7%
26Claude 2.1· Stanford HELM specialized suites data docs90.3%
27DBRX-instruct· Stanford HELM specialized suites data docs90.3%
28GPT-4o mini· Stanford HELM specialized suites data docs90.3%
29Phi-3 Medium 4K Instruct· Stanford HELM specialized suites data docs90.3%
30Qwen1.5-110B-Chat· Stanford HELM specialized suites data docs90.3%
31Qwen1.5-72B· Stanford HELM specialized suites data docs89.9%
32Claude 3 Sonnet· Stanford HELM specialized suites data docs89.5%
33mixtral-8x22b· Stanford HELM specialized suites data docs89.5%
34gemma2-27b· Stanford HELM specialized suites data docs89%
35Mistral Largesetting not named by the source· Stanford HELM specialized suites data docs88.6%
36Nova Litev1· Stanford HELM specialized suites data docs88.6%
37Claude Haiku 3.5· Stanford HELM specialized suites data docs88.2%
38Llama 2 70b· Stanford HELM specialized suites data docs88.2%
39Claude 3 Haiku· Stanford HELM specialized suites data docs87.8%
40Claude Instant 1.2· Stanford HELM specialized suites data docs87.8%
41Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs87.8%
42gemini-1.5-flash-002· Stanford HELM specialized suites data docs86.9%
43gemma-2-9b· Stanford HELM specialized suites data docs86.5%
44gemma-7b· Stanford HELM specialized suites data docs85.7%
45GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs85.7%
46Mistral Small· Stanford HELM specialized suites data docs85.7%
47Open Mistral Nemo· Stanford HELM specialized suites data docs84.8%
48Phi-3 Small 8K Instruct· Stanford HELM specialized suites data docs84.8%
49Command R+· Stanford HELM specialized suites data docs84%
50Jamba 1.5 Mini· Stanford HELM specialized suites data docs84%

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