Skip to content

Benchmark

HELM MMLU: Computer Security - EM

Scores are percentages. Higher is better.

Models measured
64
Scores recorded
68
Sources
1
IQ weight
1.73

Score distribution

78.9%
Mean
80%
Median
8%
Std dev
72%
10th percentile
85%
90th percentile
30%
Min
30%89%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.73 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 docs89%
2Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs87%
3DeepSeek V3· Stanford HELM specialized suites data docs86%
4GPT 4· Stanford HELM specialized suites data docs86%
5GPT 4 1106 Preview· Stanford HELM specialized suites data docs86%
6Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs86%
7Claude 3 Opus· Stanford HELM specialized suites data docs85%
8gemini-1.5-pro-002· Stanford HELM specialized suites data docs85%
9GPT 4osetting not named by the source· Stanford HELM specialized suites data docs85%
10GPT 4osetting not named by the source· Stanford HELM specialized suites data docs85%
11GPT-4o mini· Stanford HELM specialized suites data docs85%
12Meta Llama 3 70b· Stanford HELM specialized suites data docs85%
13Qwen2 72B Instruct· Stanford HELM specialized suites data docs85%
14Claude Haiku 3.5· Stanford HELM specialized suites data docs84%
15gemma2-27b· Stanford HELM specialized suites data docs84%
16mixtral-8x22b· Stanford HELM specialized suites data docs84%
17Nova Prov1· Stanford HELM specialized suites data docs84%
18Qwen1.5-14B· Stanford HELM specialized suites data docs84%
19Snowflake Arctic Instruct· Stanford HELM specialized suites data docs84%
20DBRX-instruct· Stanford HELM specialized suites data docs83%
21Gemini 2.0 Flash Experimental· Stanford HELM specialized suites data docs83%
22gemini-1.5-pro-001· Stanford HELM specialized suites data docs83%
23GPT-4 Turbo· Stanford HELM specialized suites data docs83%
24Yi-34B· Stanford HELM specialized suites data docs83%
25Qwen1.5-110B-Chat· Stanford HELM specialized suites data docs82%
26Claude 2.1· Stanford HELM specialized suites data docs81%
27gemma-2-9b· Stanford HELM specialized suites data docs81%
28GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs81%
29Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs81%
30Llama-3.2-90B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs81%
31Mistral Large2407· Stanford HELM specialized suites data docs81%
32Open Mistral Nemo· Stanford HELM specialized suites data docs81%
33Qwen1.5-72B· Stanford HELM specialized suites data docs81%
34Jamba 1.5 Large· Stanford HELM specialized suites data docs80%
35Llama 3.3 70B Instruct Turbo· Stanford HELM specialized suites data docs80%
36Meta Llama 3 8b· Stanford HELM specialized suites data docs80%
37Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs80%
38Mistral Largesetting not named by the source· Stanford HELM specialized suites data docs80%
39Claude 3 Haiku· Stanford HELM specialized suites data docs79%
40Claude 3 Sonnet· Stanford HELM specialized suites data docs79%
41DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs79%
42gemini-1.5-flash-001· Stanford HELM specialized suites data docs79%
43Nova Litev1· Stanford HELM specialized suites data docs79%
44Phi-3 Medium 4K Instruct· Stanford HELM specialized suites data docs79%
45Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs79%
46Command R+· Stanford HELM specialized suites data docs78%
47GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs78%
48Llama 2 70b· Stanford HELM specialized suites data docs77%
49Mistral Small· Stanford HELM specialized suites data docs77%
50Nova Microv1· Stanford HELM specialized suites data docs77%

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