Skip to content

Benchmark

HELM MMLU: Professional Accounting - EM

Scores are percentages. Higher is better.

Models measured
64
Scores recorded
68
Sources
1
IQ weight
1.94

Score distribution

59.2%
Mean
61.3%
Median
12.7%
Std dev
44.7%
10th percentile
73.4%
90th percentile
22.3%
Min
22.3%84%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.94 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 docs84%
2DeepSeek V3· Stanford HELM specialized suites data docs82.6%
3Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs81.6%
4GPT 4osetting not named by the source· Stanford HELM specialized suites data docs80.1%
5Claude 3 Opus· Stanford HELM specialized suites data docs79.1%
6gemini-1.5-pro-002· Stanford HELM specialized suites data docs78.7%
7GPT 4osetting not named by the source· Stanford HELM specialized suites data docs77%
8Gemini 2.0 Flash Experimental· Stanford HELM specialized suites data docs73.4%
9Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs72.7%
10Qwen2 72B Instruct· Stanford HELM specialized suites data docs70.9%
11text-unicorn· Stanford HELM specialized suites data docs70.6%
12GPT 4· Stanford HELM specialized suites data docs70.2%
13Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs70.2%
14GPT 4 1106 Preview· Stanford HELM specialized suites data docs68.4%
15GPT-4 Turbo· Stanford HELM specialized suites data docs68.1%
16Nova Prov1· Stanford HELM specialized suites data docs68.1%
17gemini-1.5-pro-001· Stanford HELM specialized suites data docs67.4%
18Yi-34B· Stanford HELM specialized suites data docs67.4%
19GPT-4o mini· Stanford HELM specialized suites data docs66.7%
20Llama 3.3 70B Instruct Turbo· Stanford HELM specialized suites data docs66.7%
21gemini-1.5-flash-001· Stanford HELM specialized suites data docs66%
22Llama-3.2-90B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs66%
23Phi-3 Medium 4K Instruct· Stanford HELM specialized suites data docs66%
24Claude 3 Sonnet· Stanford HELM specialized suites data docs65.6%
25Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs65.6%
26Meta Llama 3 70b· Stanford HELM specialized suites data docs64.9%
27Mistral Large2407· Stanford HELM specialized suites data docs64.5%
28Jamba 1.5 Large· Stanford HELM specialized suites data docs64.2%
29Qwen1.5-110B-Chat· Stanford HELM specialized suites data docs63.5%
30Phi-3 Small 8K Instruct· Stanford HELM specialized suites data docs63.1%
31Qwen1.5-72B· Stanford HELM specialized suites data docs63.1%
32gemini-1.5-flash-002· Stanford HELM specialized suites data docs62.1%
33mixtral-8x22b· Stanford HELM specialized suites data docs62.1%
34Claude 2.1· Stanford HELM specialized suites data docs61.3%
35Nova Litev1· Stanford HELM specialized suites data docs60.3%
36Claude Haiku 3.5· Stanford HELM specialized suites data docs58.9%
37Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs58.9%
38DBRX-instruct· Stanford HELM specialized suites data docs58.5%
39Claude Instant 1.2· Stanford HELM specialized suites data docs58.2%
40gemma2-27b· Stanford HELM specialized suites data docs57.8%
41DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs57.1%
42Llama 2 70b· Stanford HELM specialized suites data docs57.1%
43Claude 3 Haiku· Stanford HELM specialized suites data docs53.9%
44GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs53.5%
45Jamba 1.5 Mini· Stanford HELM specialized suites data docs53.5%
46Open Mistral Nemo· Stanford HELM specialized suites data docs52.8%
47Snowflake Arctic Instruct· Stanford HELM specialized suites data docs52.5%
48Nova Microv1· Stanford HELM specialized suites data docs52.1%
49Command R Plus· Stanford HELM specialized suites data docs51.4%
50gemma-2-9b· Stanford HELM specialized suites data docs51.4%

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