Compare AI models

Specifications
Parameters
16B/236B14.3B
Context window
128k32k
Benchmarks
HumanEval
90.2%34.2%
MMLU
79.2%62.5%
GSM8K
94.9%61.5%
Overview
CompanyDeepSeekQwen
Release dateJun 14 2024Mar 28 2024
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

DeepSeek Coder V2 leads Qwen1.5-MoE on 3 of the 3 benchmarks they both report (HumanEval, MMLU, GSM8K). Qwen1.5-MoE shipped 78 days before DeepSeek Coder V2, so benchmark comparisons should account for the intervening progress.

DeepSeek Coder V2 has 16B/236B parameters, while Qwen1.5-MoE has 14.3B. Context windows are 128k (DeepSeek Coder V2) vs 32k (Qwen1.5-MoE).

On HumanEval, DeepSeek Coder V2 leads at 90.2% vs Qwen1.5-MoE at 34.2%. On MMLU, DeepSeek Coder V2 leads at 79.2% vs Qwen1.5-MoE at 62.5%. On GSM8K, DeepSeek Coder V2 leads at 94.9% vs Qwen1.5-MoE at 61.5%.