Compare AI models

Specifications
Parameters
16B/236B14B
Context window
128k2k
Benchmarks
HumanEval
90.2%32.3%
MATH
75.7%24.8%
MMLU
79.2%66.3%
GSM8K
94.9%61.3%
Overview
CompanyDeepSeekQwen
Release dateJun 14 2024Sep 25 2023
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

DeepSeek Coder V2 leads Qwen-14B on 4 of the 4 benchmarks they both report (HumanEval, MATH, MMLU, GSM8K). Qwen-14B shipped 263 days before DeepSeek Coder V2, so benchmark comparisons should account for the intervening progress.

DeepSeek Coder V2 has 16B/236B parameters, while Qwen-14B has 14B. Context windows are 128k (DeepSeek Coder V2) vs 2k (Qwen-14B).

On HumanEval, DeepSeek Coder V2 leads at 90.2% vs Qwen-14B at 32.3%. On MATH, DeepSeek Coder V2 leads at 75.7% vs Qwen-14B at 24.8%. On MMLU, DeepSeek Coder V2 leads at 79.2% vs Qwen-14B at 66.3%. On GSM8K, DeepSeek Coder V2 leads at 94.9% vs Qwen-14B at 61.3%.