Compare AI models

Specifications
Parameters
671B141B
Context window
128k64k
API pricing
Cheapest input
$0.256DeepInfra$2.00Mistral
Cheapest output
$0.712DeepInfra$6.00Mistral
$103StreamLake$700Mistral · zdr
Benchmarks
MMLU
88.5%77.75%
Overview
CompanyDeepSeekMistral
Release dateDec 26 2024Apr 10 2024
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

DeepSeek V3 leads Mixtral 8×22B on 1 of the 1 benchmark they both report (MMLU). Mixtral 8×22B shipped 260 days before DeepSeek V3, so benchmark comparisons should account for the intervening progress.

DeepSeek V3 has 671B parameters, while Mixtral 8×22B has 141B. Context windows are 128k (DeepSeek V3) vs 64k (Mixtral 8×22B).

On MMLU, DeepSeek V3 leads at 88.5% vs Mixtral 8×22B at 77.75%.