Compare AI models

Specifications
Parameters
16B14.3B
Context window
32k32k
Benchmarks
HumanEval
57.3%34.2%
MMLU
55.7%62.5%
GSM8K
72%61.5%
Overview
CompanyDeepSeekQwen
Release dateMay 16 2024Mar 28 2024
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

DeepSeek V2 Lite leads Qwen1.5-MoE on 2 of the 3 benchmarks they both report (HumanEval, MMLU, GSM8K). Qwen1.5-MoE shipped 49 days before DeepSeek V2 Lite, so benchmark comparisons should account for the intervening progress.

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

On HumanEval, DeepSeek V2 Lite leads at 57.3% vs Qwen1.5-MoE at 34.2%. On MMLU, Qwen1.5-MoE leads at 62.5% vs DeepSeek V2 Lite at 55.7%. On GSM8K, DeepSeek V2 Lite leads at 72% vs Qwen1.5-MoE at 61.5%.