Chat completions API
Fully compatible with OpenAI format
DeepSeek V4 with OpenAI Codex Migration
A practical guide for migrating from OpenAI Codex/OpenAI API to DeepSeek V4 using OpenAI-compatible endpoints.
v1 Compatible
OpenAI API format
Code Focused
Optimized for coding
Cost Effective
Competitive pricing
Migration note: OpenAI Codex has been deprecated. This guide helps transition existing code to DeepSeek V4's OpenAI-compatible endpoint. Always test thoroughly before production use.
Why migrate from OpenAI Codex
Code-focused AI with OpenAI compatibility
OpenAI Codex was deprecated and integrated into GPT models. DeepSeek V4 offers a code-optimized alternative with an OpenAI-compatible API endpoint, making migration straightforward for most existing integrations.
| Feature | OpenAI Codex | DeepSeek V4 |
|---|---|---|
| Code Generation | Deprecated | Active, optimized |
| API Format | OpenAI API | OpenAI-compatible |
| Model Access | Limited availability | Public API access |
| Pricing | Premium tiers | Competitive rates |
Requirements
Existing Code
Codebase that previously used OpenAI Codex or OpenAI API
Required
API Provider
A DeepSeek Platform API key for OpenAI-compatible endpoint
Required
Runtime
Node.js 18+ or Python 3.8+ environment for API calls
Node.js 18+ / Python 3.8+
Migration Steps
1
Get DeepSeek OpenAI-compatible credentials
Obtain API key for OpenAI-compatible endpoint
- Create a DeepSeek Platform account
- Generate an API key from the dashboard
- Note the OpenAI-compatible base URL: https://api.deepseek.com/v1
- Store credentials securely (environment variables)
2
Update API configuration
Modify your OpenAI/Codex API calls to use DeepSeek endpoint
- Replace base URL with https://api.deepseek.com/v1
- Update API key to DeepSeek Platform key
- Map model names to DeepSeek equivalents
- Test with a simple completion request
3
Verify functionality
Run smoke tests to ensure core features work
- Test code generation/editing workflows
- Verify streaming responses work correctly
- Check function/tool calling if used
- Review error handling and rate limits
Configuration Reference
# OpenAI-compatible configuration
base_url: "https://api.deepseek.com/v1"
api_key: "<your-deepseek-api-key>"
# Model mapping
default_model: "deepseek-v4-pro[1m]"
fast_model: "deepseek-v4-flash"
# Example cURL
curl https://api.deepseek.com/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $DEEPSEEK_API_KEY" \
-d '{
"model": "deepseek-v4-pro[1m]",
"messages": [{"role": "user", "content": "Write a Python function"}]
}'Replace OpenAI credentials with DeepSeek equivalents. Keep API keys secure.
API Migration Comparison
| Use Case | OpenAI Format | DeepSeek Notes |
|---|---|---|
| Code Generation | POST /v1/completions or /v1/chat/completions | Change model to deepseek-v4-pro[1m] for best quality |
| Code Editing | Chat completions with edit instructions | Flash model for faster edits, Pro for complex refactors |
| Function Calling | tools/functions in chat completions | Verify function schema compatibility |
Troubleshooting
Authentication Failed
Verify API key is valid DeepSeek key, not OpenAI key. Check base URL is https://api.deepseek.com/v1
Model Not Found
Use documented DeepSeek model IDs: deepseek-v4-pro[1m] or deepseek-v4-flash. Do not use OpenAI model names like gpt-4
Different Response Format
DeepSeek's OpenAI compatibility is high but not 100%. Review compatibility table for edge cases
Rate Limit Errors
DeepSeek has different rate limits. Check account dashboard and implement exponential backoff
Compatibility status
What's supported by the OpenAI-compatible endpoint
DeepSeek's OpenAI-compatible endpoint covers the most commonly used APIs. Review the compatibility status before migration.
Completions API
Legacy completions endpoint supported
Function/tool calling
Standard tool calling format supported
Streaming responses
Server-sent events work as expected
Embeddings API
Use separate embeddings endpoint if needed
Image generation
Not supported via OpenAI compatibility
Audio/speech
Not supported via OpenAI compatibility
Secure your API integration
When migrating from OpenAI to DeepSeek, review your security practices. API keys should never be committed to version control, and rate limiting should be implemented to control costs.
- Store API keys in environment variables or secret managers
- Implement request rate limiting and cost monitoring
- Use separate keys for development and production
- Rotate keys periodically and after suspected exposure
- Audit code for hardcoded credentials before deployment
Optimize for cost and quality
DeepSeek V4 offers Pro and Flash models for different use cases. Use Pro for complex reasoning and high-quality output, Flash for faster responses on simpler tasks.
Monitor your usage through the DeepSeek dashboard and implement caching where appropriate to reduce redundant API calls.
Open the DeepSeek API pricing guideCommon Workflows
Code Generation
Generate code from natural language descriptions
- • Send prompt to chat completions
- • Specify desired language/framework
- • Review and test generated code
Code Refactoring
Improve existing code structure and quality
- • Provide code context
- • Describe refactoring goals
- • Apply suggested changes
Debugging Assistance
Get help identifying and fixing bugs
- • Share error messages and code
- • Describe expected behavior
- • Implement suggested fixes
Frequently Asked Questions
Is OpenAI Codex still available?
OpenAI Codex has been deprecated and integrated into GPT models. This guide helps migrate to DeepSeek V4 as an alternative for code-focused AI tasks.
What's the difference between OpenAI and DeepSeek endpoints?
DeepSeek provides an OpenAI-compatible endpoint at https://api.deepseek.com/v1. The request/response format is largely the same, but model IDs and some capabilities differ.
Can I use the same code with minimal changes?
Yes, for most use cases. Update the base URL, API key, and model IDs. Some advanced features may need adjustment per the compatibility table.
Which DeepSeek model should I use?
Use deepseek-v4-pro[1m] for complex reasoning and high-quality output. Use deepseek-v4-flash for faster responses and simpler tasks.
Are function calling and tools supported?
Yes, DeepSeek's OpenAI-compatible endpoint supports function/tool calling. Verify your function schemas work with the compatibility layer.
What about image and multimodal capabilities?
DeepSeek's OpenAI compatibility focuses on text/code. Check the current compatibility table for multimodal support status.