Skip to content

Benchmark

HELM MMLU: College Chemistry - EM

Scores are percentages. Higher is better.

Models measured
64
Scores recorded
68
Sources
1
IQ weight
1.76

Score distribution

51.6%
Mean
52%
Median
8.1%
Std dev
39%
10th percentile
62%
90th percentile
27%
Min
27%67%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.76 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 docs67%
2DeepSeek V3· Stanford HELM specialized suites data docs65%
3Qwen2 72B Instruct· Stanford HELM specialized suites data docs65%
4Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs63%
5Nova Prov1· Stanford HELM specialized suites data docs63%
6gemini-1.5-pro-001· Stanford HELM specialized suites data docs62%
7Qwen1.5-110B-Chat· Stanford HELM specialized suites data docs62%
8Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs62%
9GPT 4osetting not named by the source· Stanford HELM specialized suites data docs61%
10Claude 3 Opus· Stanford HELM specialized suites data docs60%
11gemini-1.5-flash-001· Stanford HELM specialized suites data docs60%
12GPT 4osetting not named by the source· Stanford HELM specialized suites data docs60%
13Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs60%
14Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs59%
15Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs59%
16text-unicorn· Stanford HELM specialized suites data docs59%
17Llama-3.2-90B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs57%
18mixtral-8x22b· Stanford HELM specialized suites data docs57%
19Gemini 2.0 Flash Experimental· Stanford HELM specialized suites data docs56%
20gemini-1.5-flash-002· Stanford HELM specialized suites data docs56%
21Meta Llama 3 70b· Stanford HELM specialized suites data docs56%
22Nova Litev1· Stanford HELM specialized suites data docs56%
23Phi-3 Medium 4K Instruct· Stanford HELM specialized suites data docs56%
24Command R Plus· Stanford HELM specialized suites data docs55%
25GPT 4· Stanford HELM specialized suites data docs55%
26Jamba 1.5 Large· Stanford HELM specialized suites data docs55%
27Llama 3.3 70B Instruct Turbo· Stanford HELM specialized suites data docs55%
28Nova Microv1· Stanford HELM specialized suites data docs54%
29Phi-3 Small 8K Instruct· Stanford HELM specialized suites data docs54%
30gemma-2-9b· Stanford HELM specialized suites data docs53%
31gemma2-27b· Stanford HELM specialized suites data docs53%
32GPT-4 Turbo· Stanford HELM specialized suites data docs53%
33Claude 3 Haiku· Stanford HELM specialized suites data docs52%
34Llama 2 70b· Stanford HELM specialized suites data docs52%
35Mistral Large2407· Stanford HELM specialized suites data docs52%
36Yi-34B· Stanford HELM specialized suites data docs52%
37Claude Haiku 3.5· Stanford HELM specialized suites data docs51%
38Claude Instant 1.2· Stanford HELM specialized suites data docs51%
39DBRX-instruct· Stanford HELM specialized suites data docs51%
40DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs51%
41GPT-4o mini· Stanford HELM specialized suites data docs51%
42Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs51%
43Claude 3 Sonnet· Stanford HELM specialized suites data docs50%
44GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs50%
45Qwen1.5-72B· Stanford HELM specialized suites data docs50%
46Claude 2.1· Stanford HELM specialized suites data docs49%
47GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs49%
48gemma-7b· Stanford HELM specialized suites data docs48%
49Meta Llama 3 8b· Stanford HELM specialized suites data docs48%
50GPT 4 1106 Preview· Stanford HELM specialized suites data docs47%

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