Compare AI models

Specifications
Parameters
236B14.3B
Context window
128k32k
Benchmarks
HumanEval
81.1%34.2%
MMLU
77.8%62.5%
GSM8K
92.2%61.5%
Overview
CompanyDeepSeekQwen
Release dateMay 6 2024Mar 28 2024
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

DeepSeek V2 leads Qwen1.5-MoE on 3 of the 3 benchmarks they both report (HumanEval, MMLU, GSM8K). Qwen1.5-MoE shipped 39 days before DeepSeek V2, so benchmark comparisons should account for the intervening progress.

DeepSeek V2 has 236B parameters, while Qwen1.5-MoE has 14.3B. Context windows are 128k (DeepSeek V2) vs 32k (Qwen1.5-MoE).

On HumanEval, DeepSeek V2 leads at 81.1% vs Qwen1.5-MoE at 34.2%. On MMLU, DeepSeek V2 leads at 77.8% vs Qwen1.5-MoE at 62.5%. On GSM8K, DeepSeek V2 leads at 92.2% vs Qwen1.5-MoE at 61.5%.