Deploying DeepSeek V3 on Akash

Introduction

DeepSeek V3 is an advanced open-source large language model (LLM) developed by DeepSeek AI. It provides high-quality natural language processing (NLP) capabilities, making it suitable for tasks such as:

Chatbots & Virtual Assistants – Powering AI-driven conversational agents.
Code Generation & Analysis – Assisting developers in writing, debugging, and optimizing code.
Content Creation – Generating human-like text for articles, reports, and creative writing.
Summarization & Translation – Improving productivity by condensing text or translating languages.

By deploying DeepSeek V3 on Akash, we leverage decentralized computing infrastructure to run the model cost-effectively, avoiding reliance on centralized cloud providers.

Prerequisites

Ensure you have the following installed on your Linux/macOS system:

Docker (Installation Guide)
Akash CLI (Installation Guide)

Provider Services CLI (Replaces the old akash CLI)

curl -sSfL https://raw.githubusercontent.com/akash-network/provider-services/main/install.sh | sh

A funded Akash wallet (Funding Guide)

Step 1: Build & Push DeepSeek V3 Docker Image

Clone the DeepSeek V3 repository and build a Docker container.

git clone https://github.com/deepseek-ai/DeepSeek-V3.git
cd DeepSeek-V3

Modify the Dockerfile as needed, then build and push the image:

docker build -t <your-dockerhub-username>/deepseek-v3 .
docker push <your-dockerhub-username>/deepseek-v3

Replace <your-dockerhub-username> with your actual Docker Hub username.

Step 2: Prepare the SDL File

Create a file called deepseek-v3.yaml with the following Service Definition Language (SDL) content:

services:
  deepseek-v3:
    image: <your-dockerhub-username>/deepseek-v3:latest
    expose:
      - port: 8000
        as: 80
        to:
          - global: true
profiles:
  compute:
    deepseek-v3:
      resources:
        cpu:
          units: 4
        memory:
          size: 16Gi
        storage:
          size: 50Gi
  placement:
    dcloud:
      attributes:
        region: us-west
      pricing:
        deepseek-v3:
          denom: uakt
          amount: 1000
deployment:
  deepseek-v3:
    dcloud:
      profile: deepseek-v3
      count: 1

Modify the image field with your Docker Hub image name.

Step 3: Deploy on Akash

1. Submit Deployment

provider-services tx deployment create deepseek-v3.yaml --from <your-wallet-name>

2. Query Deployment Status

provider-services query deployment list --owner <your-wallet-address>

Once a provider accepts the deployment, you can retrieve logs using:

provider-services query deployment logs --dseq <deployment-sequence-number>

3. Access DeepSeek V3

Once deployed, you can find your service URL with:

provider-services query deployment get --dseq <deployment-sequence-number>

Navigate to the URL in your browser or use curl:

curl http://<your-service-url>

Step 4: Manage & Delete Deployment

Update Deployment

Modify deepseek-v3.yaml and run:

provider-services tx deployment update deepseek-v3.yaml --from <your-wallet-name>

Close Deployment

provider-services tx deployment close --dseq <deployment-sequence-number> --from <your-wallet-name>

Conclusion

You have successfully deployed DeepSeek V3 on Akash! This setup allows you to run LLM workloads on a decentralized cloud, reducing costs and improving scalability.