🔄 Migrating from OpenAI to DeepSeek: Complete Guide

Last updated: July 29, 2026 · 8 min read · Python, Node.js, cURL examples

Save 80-98% on your AI API bill in 5 minutes
Drop-in replacement for OpenAI. Same SDK, same format, lower price.
Get API Key →

The 2-Line Migration

If you're using the official OpenAI SDK, migration is literally two changes:

- client = OpenAI(api_key="sk-...")
+ client = OpenAI(api_key="your-aitoken-key", base_url="https://eaf9553505eeb8f5-115-190-107-107.serveousercontent.com/v1")

- model="gpt-4o"
+ model="deepseek-v4"

Everything else — your prompts, parsing logic, streaming code, function calling — works the same.

Step-by-Step Migration

1 Get Your API Key

2 Choose Your Model

`deepseek-v4`	→ Replace GPT-4o / Claude Sonnet
`deepseek-v3`	→ Replace GPT-4o-mini / cheaper alternative
`deepseek-r1`	→ Replace o1/o3 for reasoning tasks
`deepseek-v4-flash`	→ Replace GPT-4o-mini for simple tasks
`qwen3.6-flash`	→ Ultra-cheap, high throughput

3 Update Your Code

Python (openai SDK)

from openai import OpenAI

# Before
client = OpenAI(api_key="sk-proj-...")

# After
client = OpenAI(
    api_key="ait-your-key-here",
    base_url="https://eaf9553505eeb8f5-115-190-107-107.serveousercontent.com/v1"
)

# Everything else stays the same
response = client.chat.completions.create(
    model="deepseek-v4",  # changed from gpt-4o
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Hello!"}
    ],
    temperature=0.7,
    max_tokens=1024,
)
print(response.choices[0].message.content)

Node.js (openai SDK)

import OpenAI from 'openai';

// Before
// const client = new OpenAI({ apiKey: 'sk-...' });

// After
const client = new OpenAI({
  apiKey: 'ait-your-key-here',
  baseURL: 'https://eaf9553505eeb8f5-115-190-107-107.serveousercontent.com/v1'
});

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

console.log(response.choices[0].message.content);

cURL

curl https://eaf9553505eeb8f5-115-190-107-107.serveousercontent.com/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer ait-your-key-here" \
  -d '{
    "model": "deepseek-v4",
    "messages": [
      {"role": "user", "content": "Hello!"}
    ]
  }'

Framework-Specific Migration

LangChain

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    model="deepseek-v4",
    base_url="https://eaf9553505eeb8f5-115-190-107-107.serveousercontent.com/v1",
    api_key="ait-your-key-here",
)
# All chains, agents, and tools work unchanged
# See our full LangChain guide →

CrewAI

from crewai import LLM, Agent

llm = LLM(
    model="deepseek-v4",
    base_url="https://eaf9553505eeb8f5-115-190-107-107.serveousercontent.com/v1",
    api_key="ait-your-key-here",
)

agent = Agent(role="Researcher", llm=llm, ...)
# All crew workflows work unchanged

LlamaIndex

from llama_index.llms.openai import OpenAI

llm = OpenAI(
    model="deepseek-v4",
    api_base="https://eaf9553505eeb8f5-115-190-107-107.serveousercontent.com/v1",
    api_key="ait-your-key-here",
)
# All RAG pipelines work unchanged

What Works Out of the Box

✅ Fully Compatible:

Chat completions (sync & streaming)
Function calling / Tool use
JSON mode (response_format)
System/user/assistant message roles
Temperature, top_p, max_tokens parameters
All OpenAI SDKs (Python, Node.js, etc.)

Known Differences

⚠️ Minor Differences to Be Aware Of:

Embeddings: Use deepseek-v4 or dedicated embedding models if needed (check availability)
Image input: Vision capabilities depend on the specific model — DeepSeek V4 supports it
Very new OpenAI features: May take 1-2 weeks to be supported after OpenAI releases them
Token counting: Tokenizer may differ slightly from tiktoken; costs are still per-token

Testing Checklist

Before going to production:

Test your core prompts — verify output quality matches expectations
Test function calling — ensure tool schemas are parsed correctly
Test streaming — verify SSE events work in your framework
Test error handling — try invalid inputs, check error format
Load test — verify latency meets your requirements
Monitor costs — compare first week's bill vs OpenAI

Cost Savings After Migration

Your Current Usage	OpenAI Cost	AI Token Hub Cost	You Save
1M tokens/day (GPT-4o)	~$75/week	~$15/week	$60/week
10M tokens/day (GPT-4o-mini)	~$42/week	~$5/week	$37/week
100K tokens/day (o1-mini)	~$21/week	~$4/week (R1)	$17/week

Ready to migrate? Start free today.
100 free requests/day. No credit card. OpenAI-compatible API.
Get API Key →