LLM Hosting Pro
← All model families

DeepSeek

Large open-weight mixture-of-experts models, known for strong reasoning and coding.

CodingReasoning

Overview

DeepSeek is a Chinese AI lab whose open-weight models, including DeepSeek-V3 and the R1 reasoning model, compete with leading proprietary models.

Its large mixture-of-experts models need multi-GPU hardware to self-host, so many people use them through hosted providers. Recent releases use the MIT licence.

Models in this family

ModelReleasedParametersContext
DeepSeek-V3.121 Aug 2025671B total, 37B active (mixture of experts)128K tokens
DeepSeek-R120 Jan 2025671B total, 37B active (mixture of experts)128K tokens
DeepSeek-V326 Dec 2024671B total, 37B active (mixture of experts)128K tokens

Where to host DeepSeek models

All hosts →
LLM hostModelsHosting plans
Amazon AWSDeepSeek-V3.1, DeepSeek-R1, DeepSeek-V3Amazon EC2, Amazon Bedrock, Amazon SageMaker
Google CloudDeepSeek-V3.1, DeepSeek-R1, DeepSeek-V3Model Garden, Google Kubernetes Engine (GKE), Compute Engine
Microsoft AzureDeepSeek-V3.1, DeepSeek-R1, DeepSeek-V3Azure Machine Learning, Azure Kubernetes Service (AKS), Azure GPU Virtual Machines

News about DeepSeek

All news →