Compare AI models

Specifications
Parameters
236B6B
Context window
128k8k
Benchmarks
HumanEval
81.1%58.5%
MATH
53.9%25.7%
MMLU
77.8%61.4%
GSM8K
92.2%72.3%
Overview
CompanyDeepSeekZ.ai
Release dateMay 6 2024Oct 27 2023
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

DeepSeek V2 leads ChatGLM3-6B on 4 of the 4 benchmarks they both report (HumanEval, MATH, MMLU, GSM8K). ChatGLM3-6B shipped 192 days before DeepSeek V2, so benchmark comparisons should account for the intervening progress.

DeepSeek V2 has 236B parameters, while ChatGLM3-6B has 6B. Context windows are 128k (DeepSeek V2) vs 8k (ChatGLM3-6B).

On HumanEval, DeepSeek V2 leads at 81.1% vs ChatGLM3-6B at 58.5%. On MATH, DeepSeek V2 leads at 53.9% vs ChatGLM3-6B at 25.7%. On MMLU, DeepSeek V2 leads at 77.8% vs ChatGLM3-6B at 61.4%. On GSM8K, DeepSeek V2 leads at 92.2% vs ChatGLM3-6B at 72.3%.