Compare AI models

Specifications
Parameters
16B—
Context window
4k—
Benchmarks
HumanEval
45.7%72%
MATH
15.2%47.9%
MMLU
47.2%81.5%
GSM8K
62.2%87.6%
Overview
CompanyDeepSeekZ.ai
Release dateJan 11 2024Jan 16 2024
AccessOpen WeightClosed
Model detailsView modelView model

Frequently asked questions

GLM-4 leads DeepSeek-MoE on 4 of the 4 benchmarks they both report (HumanEval, MATH, MMLU, GSM8K). DeepSeek-MoE shipped 5 days before GLM-4, so benchmark comparisons should account for the intervening progress.

DeepSeek-MoE is open weight, while GLM-4 is closed.

On HumanEval, GLM-4 leads at 72% vs DeepSeek-MoE at 45.7%. On MATH, GLM-4 leads at 47.9% vs DeepSeek-MoE at 15.2%. On MMLU, GLM-4 leads at 81.5% vs DeepSeek-MoE at 47.2%. On GSM8K, GLM-4 leads at 87.6% vs DeepSeek-MoE at 62.2%.