Skip to content

Benchmark

HELM MMLU: World Religions - EM

Scores are percentages. Higher is better.

Models measured
64
Scores recorded
68
Sources
1
IQ weight
2.11

Score distribution

84%
Mean
86%
Median
8.9%
Std dev
78.4%
10th percentile
90.1%
90th percentile
23.4%
Min
23.4%92.4%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.11 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 docs92.4%
2gemma2-27b· Stanford HELM specialized suites data docs92.4%
3DeepSeek V3· Stanford HELM specialized suites data docs91.2%
4Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs90.6%
5Meta Llama 3 70b· Stanford HELM specialized suites data docs90.6%
6Claude 3 Opus· Stanford HELM specialized suites data docs90.1%
7Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs90.1%
8Llama-3.2-90B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs90.1%
9mixtral-8x22b· Stanford HELM specialized suites data docs90.1%
10Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs90.1%
11Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs89.5%
12gemini-1.5-pro-002· Stanford HELM specialized suites data docs88.9%
13GPT 4osetting not named by the source· Stanford HELM specialized suites data docs88.9%
14gemini-1.5-flash-001· Stanford HELM specialized suites data docs88.3%
15GPT 4osetting not named by the source· Stanford HELM specialized suites data docs88.3%
16Llama 3.3 70B Instruct Turbo· Stanford HELM specialized suites data docs88.3%
17Mistral Largesetting not named by the source· Stanford HELM specialized suites data docs88.3%
18Qwen1.5-72B· Stanford HELM specialized suites data docs88.3%
19GPT 4· Stanford HELM specialized suites data docs87.7%
20Nova Prov1· Stanford HELM specialized suites data docs87.7%
21text-unicorn· Stanford HELM specialized suites data docs87.7%
22Yi-34B· Stanford HELM specialized suites data docs87.7%
23Claude 3 Haiku· Stanford HELM specialized suites data docs87.1%
24Claude 3 Sonnet· Stanford HELM specialized suites data docs87.1%
25DBRX-instruct· Stanford HELM specialized suites data docs87.1%
26Nova Litev1· Stanford HELM specialized suites data docs87.1%
27Qwen1.5-110B-Chat· Stanford HELM specialized suites data docs87.1%
28Claude Haiku 3.5· Stanford HELM specialized suites data docs86.5%
29DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs86.5%
30gemini-1.5-flash-002· Stanford HELM specialized suites data docs86.5%
31Jamba 1.5 Large· Stanford HELM specialized suites data docs86.5%
32Mistral Large2407· Stanford HELM specialized suites data docs86.5%
33Phi-3 Medium 4K Instruct· Stanford HELM specialized suites data docs86.5%
34gemma-2-9b· Stanford HELM specialized suites data docs86%
35GPT-4o mini· Stanford HELM specialized suites data docs86%
36Claude 2.1· Stanford HELM specialized suites data docs85.4%
37gemini-1.5-pro-001· Stanford HELM specialized suites data docs85.4%
38GPT 4 1106 Preview· Stanford HELM specialized suites data docs85.4%
39Llama 2 70b· Stanford HELM specialized suites data docs85.4%
40Snowflake Arctic Instruct· Stanford HELM specialized suites data docs85.4%
41GPT-4 Turbo· Stanford HELM specialized suites data docs84.8%
42Mistral Small· Stanford HELM specialized suites data docs84.8%
43Qwen2 72B Instruct· Stanford HELM specialized suites data docs84.8%
44Command R Plus· Stanford HELM specialized suites data docs84.2%
45gemma-7b· Stanford HELM specialized suites data docs84.2%
46GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs84.2%
47Jamba 1.5 Mini· Stanford HELM specialized suites data docs84.2%
48Qwen1.5-14B· Stanford HELM specialized suites data docs84.2%
49GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs83.6%
50Yi-6B· Stanford HELM specialized suites data docs83.6%

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