Compare AI models

Specifications
Parameters
16B14B
Context window
4k2k
Benchmarks
HumanEval
45.7%32.3%
MATH
15.2%24.8%
MMLU
47.2%66.3%
GSM8K
62.2%61.3%
Overview
CompanyDeepSeekQwen
Release dateJan 11 2024Sep 25 2023
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

DeepSeek-MoE and Qwen-14B are evenly matched across the 4 benchmarks they both report (HumanEval, MATH, MMLU, GSM8K). Qwen-14B shipped 108 days before DeepSeek-MoE, so benchmark comparisons should account for the intervening progress.

DeepSeek-MoE has 16B parameters, while Qwen-14B has 14B. Context windows are 4k (DeepSeek-MoE) vs 2k (Qwen-14B).

On HumanEval, DeepSeek-MoE leads at 45.7% vs Qwen-14B at 32.3%. On MATH, Qwen-14B leads at 24.8% vs DeepSeek-MoE at 15.2%. On MMLU, Qwen-14B leads at 66.3% vs DeepSeek-MoE at 47.2%. On GSM8K, DeepSeek-MoE leads at 62.2% vs Qwen-14B at 61.3%.