This article is translated from the Chinese original.

Getting the DeepSeek API and Several Ways to Use It

1. Get an NVIDIA API key

1. Register an account

Entry point: NVIDIA NIM platform DeepSeek-R1 page

Account types:

Personal email: limit of 1000 requests per day (Gmail / Outlook recommended)
Business email: limit of 5000 requests per day (any business-domain email, no verification required)

Registration process:

Click Sign In in the top-right corner and log in with a Google account.
Fill out the registration form. (The company information can be made up.)
After submission, you will immediately receive an API key.

2. Key management

Where to find it: Dashboard -> Get API Keys -> nvapi-xxxx

Proxy base URL:

https://integrate.api.nvidia.com/v1

Technical details:

Full non-distilled 671B parameter model
Supports 16k context
Average response speed: 2 to 3 seconds per request

2. Local client setup with NextChat

1. Download the client

Supported platforms: Windows / macOS / Linux

Download: GitHub Releases

2. Configure the parameters

Required fields:

API Base URL: https://integrate.api.nvidia.com/v1
API Key: nvapi-xxxx
Custom Model ID: deepseek-ai/deepseek-r1

Advanced settings:

- Temperature: 0.7 recommended (balances creativity and accuracy)
- Max Tokens: 2048 recommended (helps avoid truncation on long output)

3. Self-hosted deployment options

Option A: Deploy on Vercel for free (recommended)

Steps:

Fork the repository: open NextChat on GitHub and click Fork in the top-right corner.
Create a Vercel project:
- Log in to Vercel
- Choose Import Project and import the forked repository

Configure environment variables:

OPENAI_API_KEY=nvapi-xxxx
BASE_URL=https://integrate.api.nvidia.com/v1
CODE=access password (for example 123456)
CUSTOM_MODELS=deepseek-ai/deepseek-r1
DEFAULT_MODEL=deepseek-ai/deepseek-r1

After deployment, you will get a domain such as your-project.vercel.app. You can also bind your own domain.
Screenshot

Option B: Local Docker deployment (reference only)

docker-compose.yml:

version: '3'
services:
  nextchat:
    image: yidadaa/chatgpt-next-web
    ports:
      - "3000:3000"
    environment:
      - OPENAI_API_KEY=nvapi-xxxx
      - BASE_URL=https://integrate.api.nvidia.com/v1
      - CUSTOM_MODELS=deepseek-ai/deepseek-r1
      - DEFAULT_MODEL=deepseek-ai/deepseek-r1
    restart: unless-stopped

Start it with:

docker-compose up -d