Compare AI models

Specifications
Parameters
671B141B
Context window
128k64k
API pricing
Cheapest input
—$2.00Mistral
Cheapest output
—$6.00Mistral
—$700Mistral · zdr
Benchmarks
MMLU
90.8%77.75%
Overview
CompanyDeepSeekMistral
Release dateJan 20 2025Apr 10 2024
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

DeepSeek R1 leads Mixtral 8×22B on 1 of the 1 benchmark they both report (MMLU). Mixtral 8×22B shipped 285 days before DeepSeek R1, so benchmark comparisons should account for the intervening progress.

DeepSeek R1 has 671B parameters, while Mixtral 8×22B has 141B. Context windows are 128k (DeepSeek R1) vs 64k (Mixtral 8×22B).

On MMLU, DeepSeek R1 leads at 90.8% vs Mixtral 8×22B at 77.75%.