Compare AI models

Specifications
Parameters
16B110B
Context window
4k32k
Benchmarks
HumanEval
45.7%52.4%
MATH
15.2%49.6%
MMLU
47.2%80.4%
GSM8K
62.2%85.4%
Overview
CompanyDeepSeekQwen
Release dateJan 11 2024Apr 25 2024
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

Qwen1.5-110B leads DeepSeek-MoE on 4 of the 4 benchmarks they both report (HumanEval, MATH, MMLU, GSM8K). DeepSeek-MoE shipped 105 days before Qwen1.5-110B, so benchmark comparisons should account for the intervening progress.

DeepSeek-MoE has 16B parameters, while Qwen1.5-110B has 110B. Context windows are 4k (DeepSeek-MoE) vs 32k (Qwen1.5-110B).

On HumanEval, Qwen1.5-110B leads at 52.4% vs DeepSeek-MoE at 45.7%. On MATH, Qwen1.5-110B leads at 49.6% vs DeepSeek-MoE at 15.2%. On MMLU, Qwen1.5-110B leads at 80.4% vs DeepSeek-MoE at 47.2%. On GSM8K, Qwen1.5-110B leads at 85.4% vs DeepSeek-MoE at 62.2%.