DEV Community

Jesse
Jesse

Posted on • Edited on

How to Use DeepSeek V4 with OpenAI SDK (No Code Changes Needed)

Why DeepSeek V4?

DeepSeek V4 Pro and Flash are some of the most capable open-weight models available today. They offer:

  • 128K context window - handle long documents and conversations
  • Competitive pricing - starting at $1/1M tokens (Flash) vs $5/1M for GPT-4o
  • OpenAI-compatible API - use your existing code with zero changes

The Setup (2 minutes)

Here's the thing most developers don't realize: you can use DeepSeek with your existing OpenAI SDK code. The only change is the base_url.

Python 

from openai import OpenAI

# Just change these two lines
client = OpenAI(
    api_key="your-api-key",
    base_url="https://api.token-china.cc/v1"  # This is the only change
)

# Everything else stays the same
response = client.chat.completions.create(
    model="deepseek-v4-pro",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Explain quantum computing in simple terms."}
    ],
    temperature=0.7,
    max_tokens=500
)

print(response.choices[0].message.content)
Enter fullscreen mode Exit fullscreen mode

Node.js / TypeScript 

import OpenAI from 'openai';

const client = new OpenAI({
  apiKey: 'your-api-key',
  baseURL: 'https://api.token-china.cc/v1',  // Only change needed
});

const response = await client.chat.completions.create({
  model: 'deepseek-v4-pro',
  messages: [{ role: 'user', content: 'Hello!' }],
});

console.log(response.choices[0].message.content);
Enter fullscreen mode Exit fullscreen mode

cURL 

curl https://api.token-china.cc/v1/chat/completions   -H "Content-Type: application/json"   -H "Authorization: Bearer YOUR_API_KEY"   -d '{
    "model": "deepseek-v4-pro",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'
Enter fullscreen mode Exit fullscreen mode

Available Models

Model Context Best For Price (per 1M tokens)
DeepSeek V4 Pro 128K Complex reasoning, code $2.00
DeepSeek V4 Flash 128K Fast responses, chat $1.00
GLM 5.1 128K General purpose $1.50
GLM 5V Turbo 128K Vision tasks $3.00

Real-World Example: Building a Chatbot

Here's a complete chatbot that uses DeepSeek V4 Pro:

from openai import OpenAI

client = OpenAI(
    api_key="your-key",
    base_url="https://api.token-china.cc/v1"
)

def chat(user_message, history=[]):
    history.append({"role": "user", "content": user_message})

    response = client.chat.completions.create(
        model="deepseek-v4-pro",
        messages=[
            {"role": "system", "content": "You are a helpful coding assistant."}
        ] + history,
        temperature=0.7,
        max_tokens=1000
    )

    assistant_message = response.choices[0].message.content
    history.append({"role": "assistant", "content": assistant_message})
    return assistant_message

# Interactive chat
while True:
    user_input = input("You: ")
    if user_input.lower() in ["quit", "exit"]:
        break
    print("Bot:", chat(user_input))
Enter fullscreen mode Exit fullscreen mode

Why Use a Gateway Instead of Direct API?

If you're in China or need reliable access to Chinese AI models, a gateway service handles:

  • No phone verification - skip the Chinese phone number requirement
  • Unified API - one key for DeepSeek, GLM, and more
  • Pay-as-you-go - no monthly commitments
  • Global access - works from anywhere

I've been using Token China for my production work. The latency is good and the pricing is transparent.

Performance Comparison

I ran some benchmarks comparing DeepSeek V4 Pro via Token China vs direct GPT-4o:

Metric DeepSeek V4 Pro GPT-4o
Latency (P50) 380ms 420ms
Latency (P95) 650ms 800ms
Cost per 1K requests $0.40 $2.50
Quality (my rating) 9/10 9.5/10

The cost savings are significant - about 84% cheaper for comparable quality.

Conclusion

Switching to DeepSeek V4 is literally a one-line change. If you're paying too much for GPT-4o, give it a try. The models are genuinely good and the cost savings are real.

What's your experience with DeepSeek? Let me know in the comments.

Top comments (0)