Compare AI models

Specifications
Parameters
16B/236B—
Context window
128k—
Benchmarks
HumanEval
90.2%72%
MATH
75.7%47.9%
MMLU
79.2%81.5%
GSM8K
94.9%87.6%
Overview
CompanyDeepSeekZ.ai
Release dateJun 14 2024Jan 16 2024
AccessOpen WeightClosed
Model detailsView modelView model

Frequently asked questions

DeepSeek Coder V2 leads GLM-4 on 3 of the 4 benchmarks they both report (HumanEval, MATH, MMLU, GSM8K). GLM-4 shipped 150 days before DeepSeek Coder V2, so benchmark comparisons should account for the intervening progress.

DeepSeek Coder V2 is open weight, while GLM-4 is closed.

On HumanEval, DeepSeek Coder V2 leads at 90.2% vs GLM-4 at 72%. On MATH, DeepSeek Coder V2 leads at 75.7% vs GLM-4 at 47.9%. On MMLU, GLM-4 leads at 81.5% vs DeepSeek Coder V2 at 79.2%. On GSM8K, DeepSeek Coder V2 leads at 94.9% vs GLM-4 at 87.6%.