Skip to content

Benchmark

HELM MMLU: High School Computer Science - EM

Scores are percentages. Higher is better.

Models measured
64
Scores recorded
68
Sources
1
IQ weight
1.85

Score distribution

79.3%
Mean
81%
Median
13.6%
Std dev
66%
10th percentile
93%
90th percentile
22%
Min
22%96%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.85 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 docs96%
2DeepSeek V3· Stanford HELM specialized suites data docs95%
3Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs95%
4Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs94%
5Mistral Large2407· Stanford HELM specialized suites data docs94%
6Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs94%
7Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs93%
8gemini-1.5-pro-001· Stanford HELM specialized suites data docs93%
9GPT 4osetting not named by the source· Stanford HELM specialized suites data docs93%
10Claude 3 Opus· Stanford HELM specialized suites data docs92%
11GPT 4osetting not named by the source· Stanford HELM specialized suites data docs92%
12Llama-3.2-90B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs92%
13Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs92%
14Nova Prov1· Stanford HELM specialized suites data docs92%
15GPT-4 Turbo· Stanford HELM specialized suites data docs91%
16Qwen2 72B Instruct· Stanford HELM specialized suites data docs91%
17gemini-1.5-flash-001· Stanford HELM specialized suites data docs90%
18gemini-1.5-flash-002· Stanford HELM specialized suites data docs90%
19GPT 4 1106 Preview· Stanford HELM specialized suites data docs90%
20Jamba 1.5 Large· Stanford HELM specialized suites data docs90%
21text-unicorn· Stanford HELM specialized suites data docs90%
22GPT-4o mini· Stanford HELM specialized suites data docs88%
23GPT 4· Stanford HELM specialized suites data docs87%
24Llama 3.3 70B Instruct Turbo· Stanford HELM specialized suites data docs87%
25Meta Llama 3 70b· Stanford HELM specialized suites data docs87%
26mixtral-8x22b· Stanford HELM specialized suites data docs87%
27Gemini 2.0 Flash Experimental· Stanford HELM specialized suites data docs85%
28Claude 3 Sonnet· Stanford HELM specialized suites data docs84%
29Nova Litev1· Stanford HELM specialized suites data docs84%
30Qwen1.5-72B· Stanford HELM specialized suites data docs84%
31Phi-3 Medium 4K Instruct· Stanford HELM specialized suites data docs83%
32Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs82%
33Yi-34B· Stanford HELM specialized suites data docs82%
34Claude 2.1· Stanford HELM specialized suites data docs81%
35Qwen1.5-110B-Chat· Stanford HELM specialized suites data docs81%
36Claude Haiku 3.5· Stanford HELM specialized suites data docs80%
37DBRX-instruct· Stanford HELM specialized suites data docs80%
38Jamba 1.5 Mini· Stanford HELM specialized suites data docs79%
39Claude 3 Haiku· Stanford HELM specialized suites data docs78%
40Mistral Small· Stanford HELM specialized suites data docs78%
41Command R Plus· Stanford HELM specialized suites data docs77%
42gemma-2-9b· Stanford HELM specialized suites data docs77%
43Qwen1.5-14B· Stanford HELM specialized suites data docs77%
44GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs76%
45Llama 2 70b· Stanford HELM specialized suites data docs76%
46Nova Microv1· Stanford HELM specialized suites data docs76%
47Claude Instant 1.2· Stanford HELM specialized suites data docs75%
48Open Mistral Nemo· Stanford HELM specialized suites data docs75%
49Phi-3 Small 8K Instruct· Stanford HELM specialized suites data docs75%
50Snowflake Arctic Instruct· Stanford HELM specialized suites data docs75%

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