How do I self-host DeepSeek-V3 / R1?

DeepSeek-V3 / R1 can be deployed using Docker Compose. The AltStack provides a production-ready configuration that you can deploy on your own server. Simply copy the configuration and run it on any machine with Docker installed.

What are the system requirements for self-hosting DeepSeek-V3 / R1?

The recommended hardware for DeepSeek-V3 / R1 is: 8GB VRAM (Distilled), 160GB+ VRAM (Full). Actual requirements may vary based on usage volume and data size.

Is DeepSeek-V3 / R1 free to self-host?

Yes, DeepSeek-V3 / R1 is open source under the MIT and free to self-host. Your only costs are the server infrastructure. A basic cloud VPS starts at around $5-10/month.

What programming language is DeepSeek-V3 / R1 built with?

DeepSeek-V3 / R1 is built with Python. It has 104,505 stars on GitHub, indicating strong community adoption and support.

Back to Self-Hosted Stacks

Deploy DeepSeek-V3 / R1

Powerful open-source models including V3 (671B) and R1 (Reasoning). Rivals GPT-4o and o1.

Quick Start

Create a directory

mkdir deepseek-self-hosted && cd deepseek-self-hosted

Create config file

Copy the content from the Config tab below into a file named docker-compose.yml

Deploy

docker compose up -d

Fast Track

Deploy DeepSeek-V3 / R1 in Minutes

Use our optimized Docker templates to deploy DeepSeek-V3 / R1 on any cloud provider. Get $200 in free credit to start on DigitalOcean.

Start with free creditDeploy on DigitalOcean

$200

Metadata

License

MIT

Language

Python

Last Updated

8/28/2025

Configuration

yaml

services:
  ollama:
    image: ollama/ollama:latest
    container_name: ollama-deepseek
    restart: unless-stopped
    ports:
      - "11434:11434"
    volumes:
      - ollama_data:/root/.ollama
    # Uncomment for GPU support:
    # deploy:
    #   resources:
    #     reservations:
    #       devices:
    #         - driver: nvidia
    #           count: all
    #           capabilities: [gpu]
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:11434"]
      interval: 30s
      timeout: 10s
      retries: 3
    entrypoint: ["/bin/sh", "-c", "ollama serve & sleep 5 && ollama pull deepseek-r1 && wait"]

volumes:
  ollama_data:

Community Notes

Help others by adding context, corrections, or warnings about this tool. Sign in to contribute.

No community notes yet. Be the first to add context.