Compare AI models

Specifications
Parameters
16B—
Context window
32k—
Benchmarks
HumanEval
57.3%72%
MATH
27.9%47.9%
MMLU
55.7%81.5%
GSM8K
72%87.6%
Overview
CompanyDeepSeekZ.ai
Release dateMay 16 2024Jan 16 2024
AccessOpen WeightClosed
Model detailsView modelView model

Frequently asked questions

GLM-4 leads DeepSeek V2 Lite on 4 of the 4 benchmarks they both report (HumanEval, MATH, MMLU, GSM8K). GLM-4 shipped 121 days before DeepSeek V2 Lite, so benchmark comparisons should account for the intervening progress.

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

On HumanEval, GLM-4 leads at 72% vs DeepSeek V2 Lite at 57.3%. On MATH, GLM-4 leads at 47.9% vs DeepSeek V2 Lite at 27.9%. On MMLU, GLM-4 leads at 81.5% vs DeepSeek V2 Lite at 55.7%. On GSM8K, GLM-4 leads at 87.6% vs DeepSeek V2 Lite at 72%.