DeepSeek API Guide: V3 & R1 Setup, Pricing, and Code Examples (2026)

Call the cheapest high-performance AI models from Python, Node.js, Cursor, LangChain, and any OpenAI-compatible tool.

DeepSeek has emerged as one of the most disruptive AI model families of 2025-2026, offering performance that rivals GPT-4o and Claude at a fraction of the cost. Whether you need a budget-friendly chatbot backend, a powerful coding assistant, or a reasoning engine for complex problems, DeepSeek delivers exceptional value.

DeepSeek Model Overview

DeepSeek offers two primary models, each designed for different use cases:

DeepSeek-V3 vs DeepSeek-R1: Which to Choose?

• DeepSeek-V3 (deepseek-chat): Use for 90% of tasks — chatbots, content generation, code writing, summarization, translation, data extraction. It is fast, affordable, and high-quality.
• DeepSeek-R1 (deepseek-reasoner): Reserve for problems that require deep step-by-step reasoning — complex math proofs, multi-constraint logic puzzles, intricate debugging, and scientific analysis. It is slower and more expensive but produces more accurate answers on hard problems.

Quick Start: Python

from openai import OpenAI

client = OpenAI(
    api_key="your-token1-us-key",
    base_url="https://discount-token.com/v1"
)

response = client.chat.completions.create(
    model="deepseek-chat",
    messages=[
        {"role": "system", "content": "You are a helpful coding assistant."},
        {"role": "user", "content": "Write a Python function to reverse a linked list."}
    ],
    max_tokens=1024
)

print(response.choices[0].message.content)
print(f"\nTokens: {response.usage.total_tokens}")

Streaming Response

stream = client.chat.completions.create(
    model="deepseek-chat",
    messages=[{"role": "user", "content": "Explain Big-O notation."}],
    stream=True
)

for chunk in stream:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)

Quick Start: Node.js

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: "your-token1-us-key",
  baseURL: "https://discount-token.com/v1",
});

const response = await client.chat.completions.create({
  model: "deepseek-chat",
  messages: [{ role: "user", content: "Explain Big-O notation simply." }],
});

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

Quick Start: curl

curl https://discount-token.com/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer your-token1-us-key" \
  -d '{
    "model": "deepseek-chat",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Using DeepSeek-R1 for Complex Reasoning

DeepSeek-R1 includes a visible chain-of-thought reasoning process. The model thinks step by step before producing the final answer, which you can see in the output:

response = client.chat.completions.create(
    model="deepseek-reasoner",
    messages=[{
        "role": "user",
        "content": "A farmer has 100 meters of fencing. What is the largest rectangular area they can enclose?"
    }],
    max_tokens=4096
)

# R1's response includes the reasoning process
print(response.choices[0].message.content)

Using DeepSeek with Popular Tools

Cursor IDE

Set Cursor's OpenAI Base URL to https://discount-token.com/v1 and enter your Token1.US API key. Select deepseek-chat as the model for affordable, high-quality code assistance that costs 90% less than GPT-4o.

LangChain

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    model="deepseek-chat",
    api_key="your-token1-us-key",
    base_url="https://discount-token.com/v1"
)

response = llm.invoke("What are the SOLID principles?")

Continue (VS Code Extension)

Add a custom model in Continue's config:

{
  "models": [{
    "title": "DeepSeek-V3",
    "provider": "openai",
    "model": "deepseek-chat",
    "apiBase": "https://discount-token.com/v1",
    "apiKey": "your-token1-us-key"
  }]
}

Dify

Add a model provider with base URL https://discount-token.com/v1 and your Token1.US API key. DeepSeek will be available as a model option in all Dify apps.

DeepSeek API Pricing: How Much Does It Cost?

DeepSeek is the cheapest high-quality AI API available. Here is what real-world usage costs:

What DeepSeek Excels At

Coding and Programming

DeepSeek-V3 performs excellently on coding benchmarks, often matching GPT-4o. It is particularly strong at:

• Python, JavaScript/TypeScript, Java, C++, Go, Rust
• Algorithm implementation and optimization
• Debugging and error fixing
• Code refactoring and documentation
• SQL query generation

Chinese Language Tasks

DeepSeek was developed by a Chinese AI lab, giving it native-level Chinese understanding. For Chinese-language applications — chatbots, content generation, sentiment analysis — DeepSeek outperforms most Western models.

Bilingual Processing

DeepSeek handles mixed Chinese-English content naturally. If your application processes both languages (e.g., translation tools, international chatbots), DeepSeek is an excellent choice.

Mathematical Reasoning

DeepSeek-R1 is specifically trained for mathematical and logical reasoning. On math benchmarks, it competes with OpenAI's o1 model at a fraction of the cost.

DeepSeek vs Other AI Models

For a full comparison, see our detailed model comparison page.

Best Practices for DeepSeek API

• Use V3 by default: It covers 90% of use cases at the lowest cost. Switch to R1 only when you need deep reasoning.
• Leverage cost savings: Route bulk processing tasks to DeepSeek-V3 instead of GPT-4o to save 90%+ on API costs.
• Handle context limits: DeepSeek-V3 supports 64K tokens of context. For longer inputs, chunk your data.
• Use streaming: DeepSeek streams efficiently, improving perceived latency for chat applications.
• Implement caching: DeepSeek is so cheap that caching may seem unnecessary, but for high-volume identical queries, it still saves money.
• Batch processing: For data labeling or classification, DeepSeek-V3 is ideal due to its low cost per token.

Troubleshooting

Frequently Asked Questions

Related Guides

• What Is an AI API Gateway?
• GPT API Access Guide
• Claude API Access Guide
• GLM API Pricing Guide
• AI Model Comparison
• Cheapest AI API Guide