Compare AI models

Specifications
Parameters
16B47B
Context window
32k32k
Benchmarks
HumanEval
57.3%40.2%
MMLU
55.7%70.6%
GSM8K
72%58.4%
Overview
CompanyDeepSeekMistral
Release dateMay 16 2024Dec 8 2023
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

DeepSeek V2 Lite leads Mixtral 8×7B on 2 of the 3 benchmarks they both report (HumanEval, MMLU, GSM8K). Mixtral 8×7B shipped 160 days before DeepSeek V2 Lite, so benchmark comparisons should account for the intervening progress.

DeepSeek V2 Lite has 16B parameters, while Mixtral 8×7B has 47B. Context windows are 32k (DeepSeek V2 Lite) vs 32k (Mixtral 8×7B).

On HumanEval, DeepSeek V2 Lite leads at 57.3% vs Mixtral 8×7B at 40.2%. On MMLU, Mixtral 8×7B leads at 70.6% vs DeepSeek V2 Lite at 55.7%. On GSM8K, DeepSeek V2 Lite leads at 72% vs Mixtral 8×7B at 58.4%.