How to Use DeepSeek in Cursor & Continue IDE (Setup Guide)

AI-powered code editors like Cursor and Continue (VS Code extension) can use custom API endpoints. This means you can use DeepSeek-v3.2 through NovAI as your coding assistant — at a fraction of the cost of GPT-4.

Setup for Cursor IDE

Cursor supports custom OpenAI-compatible endpoints. Here’s how to configure it:

1. Open Cursor Settings (Cmd+, or Ctrl+,)
2. Go to Models section
3. Add a new model with these settings:

Model Name: deepseek-v3.2
API Key: your-novai-api-key
Base URL: https://aiapi-pro.com/v1

Now you can select DeepSeek-v3.2 from the model dropdown in Cursor and use it for code completion, chat, and editing.

Setup for Continue (VS Code Extension)

Add this to your ~/.continue/config.json:

{
  "models": [
    {
      "title": "DeepSeek-v3.2",
      "provider": "openai",
      "model": "deepseek-v3.2",
      "apiKey": "your-novai-api-key",
      "apiBase": "https://aiapi-pro.com/v1"
    },
    {
      "title": "GLM-Flash (Free)",
      "provider": "openai",
      "model": "glm-4.6v-flash",
      "apiKey": "your-novai-api-key",
      "apiBase": "https://aiapi-pro.com/v1"
    }
  ]
}

Why DeepSeek for Coding?

DeepSeek-v3.2 scores 89.4% on HumanEval, nearly matching GPT-4o’s 90.2%. But at $0.20/1M input tokens vs GPT-4o’s $2.50, you save over 90% on costs.

For an average coding session generating ~50K tokens of context per query and ~2K tokens of output, that’s roughly $0.0008 per query with DeepSeek vs $0.014 with GPT-4o.

Pro Tip: Use GLM-Flash for Free Practice

If you’re just learning or experimenting, use glm-4.6v-flash as your model — it’s completely free. Switch to DeepSeek when you need higher quality for production code.