Skip to content

Benchmark

HELM MMLU: International Law - EM

Scores are percentages. Higher is better.

Models measured
64
Scores recorded
68
Sources
1
IQ weight
1.94

Score distribution

85.1%
Mean
88.4%
Median
9.9%
Std dev
75.2%
10th percentile
93.4%
90th percentile
30.6%
Min
30.6%95.9%
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 docs95.9%
2Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs95.9%
3DeepSeek V3· Stanford HELM specialized suites data docs95%
4Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs95%
5GPT 4osetting not named by the source· Stanford HELM specialized suites data docs94.2%
6GPT-4 Turbo· Stanford HELM specialized suites data docs94.2%
7GPT 4osetting not named by the source· Stanford HELM specialized suites data docs93.4%
8Llama-3.2-90B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs93.4%
9Phi-3 Medium 4K Instruct· Stanford HELM specialized suites data docs93.4%
10GPT 4 1106 Preview· Stanford HELM specialized suites data docs92.6%
11GPT-4o mini· Stanford HELM specialized suites data docs92.6%
12Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs92.6%
13Mistral Large2407· Stanford HELM specialized suites data docs92.6%
14gemini-1.5-pro-001· Stanford HELM specialized suites data docs91.7%
15gemini-1.5-pro-002· Stanford HELM specialized suites data docs91.7%
16GPT 4· Stanford HELM specialized suites data docs91.7%
17mixtral-8x22b· Stanford HELM specialized suites data docs91.7%
18Claude 3 Sonnet· Stanford HELM specialized suites data docs90.9%
19Qwen1.5-72B· Stanford HELM specialized suites data docs90.9%
20text-unicorn· Stanford HELM specialized suites data docs90.9%
21Yi-34B· Stanford HELM specialized suites data docs90.9%
22Claude 2.1· Stanford HELM specialized suites data docs90.1%
23Claude 3 Haiku· Stanford HELM specialized suites data docs90.1%
24Claude 3 Opus· Stanford HELM specialized suites data docs90.1%
25gemini-1.5-flash-001· Stanford HELM specialized suites data docs90.1%
26Meta Llama 3 70b· Stanford HELM specialized suites data docs90.1%
27Nova Prov1· Stanford HELM specialized suites data docs90.1%
28Jamba 1.5 Mini· Stanford HELM specialized suites data docs89.3%
29Qwen2 72B Instruct· Stanford HELM specialized suites data docs89.3%
30Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs89.3%
31Claude Haiku 3.5· Stanford HELM specialized suites data docs88.4%
32DBRX-instruct· Stanford HELM specialized suites data docs88.4%
33Jamba 1.5 Large· Stanford HELM specialized suites data docs88.4%
34Llama 3.3 70B Instruct Turbo· Stanford HELM specialized suites data docs88.4%
35Qwen1.5-110B-Chat· Stanford HELM specialized suites data docs87.6%
36Llama 2 70b· Stanford HELM specialized suites data docs86.8%
37Mistral Largesetting not named by the source· Stanford HELM specialized suites data docs86.8%
38Qwen2.5 7B Instruct Turbo· Stanford HELM specialized suites data docs86%
39Claude Instant 1.2· Stanford HELM specialized suites data docs85.1%
40DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs85.1%
41Phi-3 Small 8K Instruct· Stanford HELM specialized suites data docs85.1%
42gemma2-27b· Stanford HELM specialized suites data docs84.3%
43GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs84.3%
44Meta Llama 3 8b· Stanford HELM specialized suites data docs84.3%
45Nova Litev1· Stanford HELM specialized suites data docs84.3%
46Nova Microv1· Stanford HELM specialized suites data docs84.3%
47Command R Plus· Stanford HELM specialized suites data docs83.5%
48gemma-2-9b· Stanford HELM specialized suites data docs83.5%
49gemma-7b· Stanford HELM specialized suites data docs83.5%
50Jamba Instruct· Stanford HELM specialized suites data docs83.5%

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