Skip to content

Benchmark

HELM MMLU: Abstract Algebra - EM

Scores are percentages. Higher is better.

Models measured
64
Scores recorded
68
Sources
1
IQ weight
1.50

Score distribution

46.7%
Mean
44%
Median
16%
Std dev
28%
10th percentile
70%
90th percentile
21%
Min
21%84%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.50 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
1DeepSeek V3· Stanford HELM specialized suites data docs84%
2gemini-1.5-pro-002· Stanford HELM specialized suites data docs82%
3Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs78%
4Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs75%
5gemini-1.5-pro-001· Stanford HELM specialized suites data docs75%
6Gemini 2.0 Flash Experimental· Stanford HELM specialized suites data docs72%
7Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs70%
8Mistral Large2407· Stanford HELM specialized suites data docs70%
9Nova Prov1· Stanford HELM specialized suites data docs69%
10Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs68%
11Qwen2 72B Instruct· Stanford HELM specialized suites data docs67%
12GPT 4osetting not named by the source· Stanford HELM specialized suites data docs66%
13Claude 3 Opus· Stanford HELM specialized suites data docs64%
14gemini-1.5-flash-002· Stanford HELM specialized suites data docs63%
15GPT 4· Stanford HELM specialized suites data docs63%
16gemini-1.5-flash-001· Stanford HELM specialized suites data docs58%
17GPT 4osetting not named by the source· Stanford HELM specialized suites data docs58%
18Qwen1.5-110B-Chat· Stanford HELM specialized suites data docs57%
19GPT-4 Turbo· Stanford HELM specialized suites data docs56%
20Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs55%
21GPT 4 1106 Preview· Stanford HELM specialized suites data docs53%
22Jamba 1.5 Large· Stanford HELM specialized suites data docs53%
23Llama-3.2-90B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs52%
24Nova Litev1· Stanford HELM specialized suites data docs52%
25text-unicorn· Stanford HELM specialized suites data docs51%
26Llama 3.3 70B Instruct Turbo· Stanford HELM specialized suites data docs50%
27Phi-3 Medium 4K Instruct· Stanford HELM specialized suites data docs50%
28Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs49%
29mixtral-8x22b· Stanford HELM specialized suites data docs48%
30Claude Haiku 3.5· Stanford HELM specialized suites data docs47%
31Mistral Largesetting not named by the source· Stanford HELM specialized suites data docs45%
32DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs44%
33Phi-3 Small 8K Instruct· Stanford HELM specialized suites data docs44%
34Qwen1.5-72B· Stanford HELM specialized suites data docs44%
35Meta Llama 3 70b· Stanford HELM specialized suites data docs43%
36Claude 3 Haiku· Stanford HELM specialized suites data docs42%
37GPT-4o mini· Stanford HELM specialized suites data docs42%
38Nova Microv1· Stanford HELM specialized suites data docs42%
39Claude 2.1· Stanford HELM specialized suites data docs40%
40gemma-2-9b· Stanford HELM specialized suites data docs40%
41gemma2-27b· Stanford HELM specialized suites data docs40%
42Qwen1.5-14B· Stanford HELM specialized suites data docs40%
43Yi-34B· Stanford HELM specialized suites data docs40%
44Claude 3 Sonnet· Stanford HELM specialized suites data docs39%
45Qwen1.5-7B· Stanford HELM specialized suites data docs39%
46GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs38%
47Claude Instant 1.2· Stanford HELM specialized suites data docs37%
48Jamba Instruct· Stanford HELM specialized suites data docs36%
49Snowflake Arctic Instruct· Stanford HELM specialized suites data docs35%
50DBRX-instruct· Stanford HELM specialized suites data docs34%

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