Skip to content

Benchmark

HELM MMLU: Clinical Knowledge - EM

Scores are percentages. Higher is better.

Models measured
64
Scores recorded
68
Sources
1
IQ weight
2.28

Score distribution

77.9%
Mean
80%
Median
10.8%
Std dev
66%
10th percentile
87.9%
90th percentile
26%
Min
26%92.8%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.28 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 docs92.8%
2Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs91.3%
3DeepSeek V3· Stanford HELM specialized suites data docs91.3%
4gemini-1.5-pro-002· Stanford HELM specialized suites data docs90.6%
5GPT 4osetting not named by the source· Stanford HELM specialized suites data docs89.4%
6GPT 4osetting not named by the source· Stanford HELM specialized suites data docs89.4%
7Claude 3 Opus· Stanford HELM specialized suites data docs87.9%
8Gemini 2.0 Flash Experimental· Stanford HELM specialized suites data docs87.9%
9Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs87.9%
10Nova Prov1· Stanford HELM specialized suites data docs87.5%
11Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs87.2%
12Qwen2 72B Instruct· Stanford HELM specialized suites data docs86.8%
13GPT 4 1106 Preview· Stanford HELM specialized suites data docs86.4%
14Mistral Large2407· Stanford HELM specialized suites data docs86.4%
15gemini-1.5-pro-001· Stanford HELM specialized suites data docs85.3%
16Jamba 1.5 Large· Stanford HELM specialized suites data docs84.9%
17GPT 4· Stanford HELM specialized suites data docs84.5%
18GPT-4o mini· Stanford HELM specialized suites data docs84.5%
19Llama-3.2-90B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs84.5%
20Meta Llama 3 70b· Stanford HELM specialized suites data docs84.5%
21Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs84.5%
22gemini-1.5-flash-001· Stanford HELM specialized suites data docs83.4%
23Qwen1.5-72B· Stanford HELM specialized suites data docs83.4%
24GPT-4 Turbo· Stanford HELM specialized suites data docs83%
25Llama 3.3 70B Instruct Turbo· Stanford HELM specialized suites data docs83%
26Phi-3 Small 8K Instruct· Stanford HELM specialized suites data docs83%
27Phi-3 Medium 4K Instruct· Stanford HELM specialized suites data docs82.6%
28Claude Haiku 3.5· Stanford HELM specialized suites data docs82.3%
29mixtral-8x22b· Stanford HELM specialized suites data docs81.9%
30gemma2-27b· Stanford HELM specialized suites data docs80.8%
31Claude 3 Sonnet· Stanford HELM specialized suites data docs80.4%
32text-unicorn· Stanford HELM specialized suites data docs80.4%
33Nova Litev1· Stanford HELM specialized suites data docs80%
34Yi-34B· Stanford HELM specialized suites data docs80%
35gemini-1.5-flash-002· Stanford HELM specialized suites data docs79.2%
36Claude 3 Haiku· Stanford HELM specialized suites data docs78.9%
37DBRX-instruct· Stanford HELM specialized suites data docs78.9%
38Claude 2.1· Stanford HELM specialized suites data docs78.5%
39DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs78.5%
40Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs78.5%
41Snowflake Arctic Instruct· Stanford HELM specialized suites data docs78.1%
42gemma-2-9b· Stanford HELM specialized suites data docs77.7%
43GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs77.7%
44Mistral Small· Stanford HELM specialized suites data docs76.6%
45Qwen1.5-110B-Chat· Stanford HELM specialized suites data docs76.6%
46GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs75.5%
47Command R+· Stanford HELM specialized suites data docs75.1%
48Meta Llama 3 8b· Stanford HELM specialized suites data docs75.1%
49Mistral Largesetting not named by the source· Stanford HELM specialized suites data docs75.1%
50Nova Microv1· Stanford HELM specialized suites data docs75.1%

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