Compare AI models

Specifications
Parameters
7B/13B/70B—
Context window
4k—
Benchmarks
HumanEval
29.9%72%
MATH
13.5%47.9%
MMLU
68.9%81.5%
GSM8K
56.8%87.6%
Overview
CompanyMetaZ.ai
Release dateJul 18 2023Jan 16 2024
AccessOpen WeightClosed
Model detailsView modelView model

Frequently asked questions

GLM-4 leads LLaMA 2 on 4 of the 4 benchmarks they both report (HumanEval, MATH, MMLU, GSM8K). LLaMA 2 shipped 182 days before GLM-4, so benchmark comparisons should account for the intervening progress.

LLaMA 2 is open weight, while GLM-4 is closed.

On HumanEval, GLM-4 leads at 72% vs LLaMA 2 at 29.9%. On MATH, GLM-4 leads at 47.9% vs LLaMA 2 at 13.5%. On MMLU, GLM-4 leads at 81.5% vs LLaMA 2 at 68.9%. On GSM8K, GLM-4 leads at 87.6% vs LLaMA 2 at 56.8%.