Skip to content

Benchmark

HELM MMLU: High School Geography - EM

Scores are percentages. Higher is better.

Models measured
64
Scores recorded
68
Sources
1
IQ weight
2.38

Score distribution

88%
Mean
90.9%
Median
10.8%
Std dev
81.8%
10th percentile
94.9%
90th percentile
22.7%
Min
22.7%96%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.38 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-001· Stanford HELM specialized suites data docs96%
2GPT 4· Stanford HELM specialized suites data docs96%
3Claude 3 Opus· Stanford HELM specialized suites data docs95.5%
4DeepSeek V3· Stanford HELM specialized suites data docs95.5%
5GPT 4osetting not named by the source· Stanford HELM specialized suites data docs95.5%
6GPT 4osetting not named by the source· Stanford HELM specialized suites data docs95.5%
7Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs95.5%
8Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs94.9%
9GPT-4 Turbo· Stanford HELM specialized suites data docs94.9%
10gemini-1.5-pro-002· Stanford HELM specialized suites data docs94.4%
11Nova Litev1· Stanford HELM specialized suites data docs94.4%
12Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs93.9%
13Jamba 1.5 Large· Stanford HELM specialized suites data docs93.9%
14Meta Llama 3 70b· Stanford HELM specialized suites data docs93.9%
15Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs93.9%
16Qwen1.5-72B· Stanford HELM specialized suites data docs93.9%
17Qwen2 72B Instruct· Stanford HELM specialized suites data docs93.9%
18gemma2-27b· Stanford HELM specialized suites data docs93.4%
19GPT 4 1106 Preview· Stanford HELM specialized suites data docs93.4%
20Llama-3.2-90B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs93.4%
21Mistral Large2407· Stanford HELM specialized suites data docs93.4%
22GPT-4o mini· Stanford HELM specialized suites data docs92.9%
23Llama 3.3 70B Instruct Turbo· Stanford HELM specialized suites data docs92.9%
24Nova Prov1· Stanford HELM specialized suites data docs92.9%
25Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs92.9%
26Claude Haiku 3.5· Stanford HELM specialized suites data docs92.4%
27mixtral-8x22b· Stanford HELM specialized suites data docs92.4%
28Phi-3 Medium 4K Instruct· Stanford HELM specialized suites data docs92.4%
29text-unicorn· Stanford HELM specialized suites data docs92.4%
30gemini-1.5-flash-001· Stanford HELM specialized suites data docs91.9%
31Qwen1.5-110B-Chat· Stanford HELM specialized suites data docs91.9%
32Claude 3 Sonnet· Stanford HELM specialized suites data docs90.9%
33Mistral Largesetting not named by the source· Stanford HELM specialized suites data docs90.9%
34Nova Microv1· Stanford HELM specialized suites data docs90.9%
35Yi-34B· Stanford HELM specialized suites data docs90.9%
36gemini-1.5-flash-002· Stanford HELM specialized suites data docs90.4%
37GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs90.4%
38Claude 2.1· Stanford HELM specialized suites data docs89.9%
39DBRX-instruct· Stanford HELM specialized suites data docs89.9%
40Phi-3 Small 8K Instruct· Stanford HELM specialized suites data docs89.9%
41Claude Instant 1.2· Stanford HELM specialized suites data docs89.4%
42gemma-2-9b· Stanford HELM specialized suites data docs89.4%
43DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs88.9%
44Gemini 2.0 Flash Experimental· Stanford HELM specialized suites data docs88.9%
45Jamba 1.5 Mini· Stanford HELM specialized suites data docs87.9%
46Llama 2 70b· Stanford HELM specialized suites data docs87.9%
47Command R Plus· Stanford HELM specialized suites data docs87.4%
48Mistral Small· Stanford HELM specialized suites data docs87.4%
49Qwen1.5-14B· Stanford HELM specialized suites data docs86.9%
50Claude 3 Haiku· Stanford HELM specialized suites data docs86.4%

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