Available Models

DeepSeek offers a range of specialized AI models optimized for different tasks and use cases.

Model Overview

Model	Context Length	Strengths	Best For
deepseek-chat	32K tokens	General conversation, reasoning	Chatbots, Q&A, general AI tasks
deepseek-coder	16K tokens	Code generation, debugging	Programming assistance, code review
deepseek-math	8K tokens	Mathematical reasoning	Math problems, calculations

DeepSeek Chat

Model Details

Model ID: deepseek-chat
Context Length: 32,768 tokens
Training Data: Diverse text from web, books, and curated sources
Capabilities: General conversation, reasoning, analysis, creative writing

Strengths

Conversational AI: Natural, engaging dialogue
Reasoning: Complex problem-solving and analysis
Multilingual: Support for multiple languages
Knowledge: Broad knowledge across domains
Safety: Built-in safety measures and content filtering

Use Cases

Customer service chatbots
Virtual assistants
Content generation
Question answering
Educational tutoring
Creative writing assistance

Example Usage

python

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_API_KEY",
    base_url="https://api.deepseek.com/v1"
)

response = client.chat.completions.create(
    model="deepseek-chat",
    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)

Performance Characteristics

Latency: ~1-3 seconds for typical responses
Throughput: High concurrent request handling
Quality: State-of-the-art response quality
Consistency: Reliable performance across different prompts

DeepSeek Coder

Model Details

Model ID: deepseek-coder
Context Length: 16,384 tokens
Training Data: Code repositories, documentation, programming resources
Capabilities: Code generation, debugging, explanation, optimization

Strengths

Multi-language Support: Python, JavaScript, Java, C++, Go, Rust, and more
Code Understanding: Analyze and explain existing code
Bug Detection: Identify and fix programming errors
Optimization: Suggest performance improvements
Documentation: Generate code comments and documentation

Supported Languages

Python
JavaScript/TypeScript
Java
C/C++
Go
Rust
PHP
Ruby
Swift
Kotlin
And many more...

Use Cases

Code generation and completion
Bug fixing and debugging
Code review and optimization
API integration assistance
Algorithm implementation
Documentation generation

Example Usage

python

response = client.chat.completions.create(
    model="deepseek-coder",
    messages=[
        {
            "role": "user", 
            "content": "Write a Python function to find the longest common subsequence of two strings."
        }
    ],
    temperature=0.2,  # Lower temperature for more deterministic code
    max_tokens=800
)

print(response.choices[0].message.content)

Code Generation Best Practices

Be specific: Provide clear requirements and constraints
Include context: Mention the programming language and framework
Specify format: Request comments, tests, or documentation
Use examples: Show input/output examples when helpful

python

# Good prompt example
prompt = """
Write a Python function that:
1. Takes a list of integers as input
2. Returns the second largest number
3. Handles edge cases (empty list, single element)
4. Include docstring and type hints
5. Add error handling for invalid inputs
"""

DeepSeek Math

Model Details

Model ID: deepseek-math
Context Length: 8,192 tokens
Training Data: Mathematical texts, problem sets, proofs, equations
Capabilities: Mathematical reasoning, problem solving, proof generation

Strengths

Mathematical Reasoning: Step-by-step problem solving
Multiple Domains: Algebra, calculus, geometry, statistics, and more
Proof Generation: Formal and informal mathematical proofs
Equation Solving: Symbolic and numerical computation
Explanation: Clear mathematical explanations

Mathematical Domains

Algebra: Linear equations, polynomials, systems
Calculus: Derivatives, integrals, limits
Geometry: Euclidean and analytic geometry
Statistics: Probability, distributions, hypothesis testing
Number Theory: Prime numbers, modular arithmetic
Discrete Math: Combinatorics, graph theory, logic

Use Cases

Educational math tutoring
Homework assistance
Mathematical research support
Engineering calculations
Financial modeling
Scientific computing

Example Usage

python

response = client.chat.completions.create(
    model="deepseek-math",
    messages=[
        {
            "role": "user",
            "content": "Solve the differential equation dy/dx = 2x + 3, with initial condition y(0) = 1. Show all steps."
        }
    ],
    temperature=0.1,  # Very low temperature for mathematical accuracy
    max_tokens=600
)

print(response.choices[0].message.content)

Mathematical Formatting

The model supports various mathematical notation formats:

LaTeX: $\int_0^1 x^2 dx = \frac{1}{3}$
Plain text: integral from 0 to 1 of x^2 dx = 1/3
Step-by-step: Detailed solution processes

Model Selection Guide

Choose DeepSeek Chat When:

Building conversational applications
Need general-purpose AI assistance
Working with diverse content types
Require multilingual support
Need creative or analytical thinking

Choose DeepSeek Coder When:

Generating or reviewing code
Building developer tools
Need programming assistance
Working on software projects
Require code explanation or documentation

Choose DeepSeek Math When:

Solving mathematical problems
Building educational applications
Need mathematical reasoning
Working with equations or proofs
Require step-by-step solutions

Model Comparison

Performance Metrics

Metric	DeepSeek Chat	DeepSeek Coder	DeepSeek Math
General Knowledge	⭐⭐⭐⭐⭐	⭐⭐⭐	⭐⭐⭐
Code Generation	⭐⭐⭐	⭐⭐⭐⭐⭐	⭐⭐
Mathematical Reasoning	⭐⭐⭐	⭐⭐	⭐⭐⭐⭐⭐
Conversation Quality	⭐⭐⭐⭐⭐	⭐⭐⭐	⭐⭐⭐
Context Understanding	⭐⭐⭐⭐⭐	⭐⭐⭐⭐	⭐⭐⭐⭐

Cost Considerations

Model	Input Cost	Output Cost	Context Length
deepseek-chat	$0.14/1M tokens	$0.28/1M tokens	32K
deepseek-coder	$0.14/1M tokens	$0.28/1M tokens	16K
deepseek-math	$0.14/1M tokens	$0.28/1M tokens	8K

Prices are subject to change. Check our pricing page for current rates.

Advanced Features

Function Calling

All models support function calling for external tool integration:

python

functions = [
    {
        "name": "calculate",
        "description": "Perform mathematical calculations",
        "parameters": {
            "type": "object",
            "properties": {
                "expression": {"type": "string", "description": "Math expression"}
            },
            "required": ["expression"]
        }
    }
]

response = client.chat.completions.create(
    model="deepseek-chat",
    messages=[{"role": "user", "content": "What's 15% of 240?"}],
    functions=functions,
    function_call="auto"
)

JSON Mode

Force structured output with JSON mode:

python

response = client.chat.completions.create(
    model="deepseek-chat",
    messages=[
        {
            "role": "system",
            "content": "You are a helpful assistant designed to output JSON."
        },
        {
            "role": "user",
            "content": "Generate a user profile for John Doe"
        }
    ],
    response_format={"type": "json_object"}
)

Streaming

All models support streaming for real-time responses:

python

stream = client.chat.completions.create(
    model="deepseek-chat",
    messages=[{"role": "user", "content": "Tell me a story"}],
    stream=True
)

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

Model Updates and Versioning

Version Naming

Models follow semantic versioning:

deepseek-chat: Latest stable version
deepseek-chat-v1.0: Specific version
deepseek-chat-preview: Preview/beta version

Update Policy

Stable models: Updated monthly with improvements
Preview models: Updated weekly with latest features
Deprecated models: 90-day notice before removal

Backward Compatibility

We maintain backward compatibility for:

API endpoints and parameters
Response formats
Core functionality

Breaking changes are introduced only in major version updates with advance notice.

Best Practices

Model-Specific Optimization

For DeepSeek Chat:

Use system messages to set context
Adjust temperature based on creativity needs
Implement conversation memory management

For DeepSeek Coder:

Be specific about programming language
Request code comments and documentation
Use lower temperature for deterministic output

For DeepSeek Math:

Request step-by-step solutions
Specify desired output format (LaTeX, plain text)
Use very low temperature for accuracy

Performance Optimization

Choose the right model: Match model capabilities to your use case
Optimize context length: Use only necessary context
Batch requests: Combine multiple queries when possible
Cache responses: Store results for repeated queries

Error Handling

python

def robust_model_call(model, messages, **kwargs):
    try:
        response = client.chat.completions.create(
            model=model,
            messages=messages,
            **kwargs
        )
        return response
    except Exception as e:
        if "model_not_found" in str(e):
            # Fallback to default model
            return client.chat.completions.create(
                model="deepseek-chat",
                messages=messages,
                **kwargs
            )
        raise e

Migration Guide

From OpenAI Models

OpenAI Model	DeepSeek Equivalent	Notes
gpt-3.5-turbo	deepseek-chat	Similar capabilities, longer context
gpt-4	deepseek-chat	Comparable performance
code-davinci-002	deepseek-coder	Specialized for coding tasks

Migration Steps

Update base URL: Change to https://api.deepseek.com/v1
Replace API key: Use your DeepSeek API key
Update model names: Use DeepSeek model identifiers
Test functionality: Verify responses meet your requirements

Future Roadmap

Upcoming Models

DeepSeek Vision: Multimodal image understanding
DeepSeek Audio: Speech and audio processing
DeepSeek Reasoning: Enhanced logical reasoning

Planned Features

Larger context windows
Faster inference speeds
Additional specialized models
Enhanced function calling

Getting Help

Documentation

Support

Ready to start using our models? Check out our Quick Start Guide or explore our API Reference.

Available Models ​

Model Overview ​

DeepSeek Chat ​

Model Details ​

Strengths ​

Use Cases ​

Example Usage ​

Performance Characteristics ​

DeepSeek Coder ​

Model Details ​

Strengths ​

Supported Languages ​

Use Cases ​

Example Usage ​

Code Generation Best Practices ​

DeepSeek Math ​

Model Details ​

Strengths ​

Mathematical Domains ​

Use Cases ​

Example Usage ​

Mathematical Formatting ​

Model Selection Guide ​

Choose DeepSeek Chat When: ​

Choose DeepSeek Coder When: ​

Choose DeepSeek Math When: ​

Model Comparison ​

Performance Metrics ​

Cost Considerations ​

Advanced Features ​

Function Calling ​

JSON Mode ​

Streaming ​

Model Updates and Versioning ​

Version Naming ​

Update Policy ​

Backward Compatibility ​

Best Practices ​

Model-Specific Optimization ​

For DeepSeek Chat: ​

For DeepSeek Coder: ​

For DeepSeek Math: ​

Performance Optimization ​

Error Handling ​

Migration Guide ​

From OpenAI Models ​

Migration Steps ​

Future Roadmap ​

Upcoming Models ​

Planned Features ​

Getting Help ​

Documentation ​

Support ​

Available Models

Model Overview

DeepSeek Chat

Model Details

Strengths

Use Cases

Example Usage

Performance Characteristics

DeepSeek Coder

Model Details

Strengths

Supported Languages

Use Cases

Example Usage

Code Generation Best Practices

DeepSeek Math

Model Details

Strengths

Mathematical Domains

Use Cases

Example Usage

Mathematical Formatting

Model Selection Guide

Choose DeepSeek Chat When:

Choose DeepSeek Coder When:

Choose DeepSeek Math When:

Model Comparison

Performance Metrics

Cost Considerations

Advanced Features

Function Calling

JSON Mode

Streaming

Model Updates and Versioning

Version Naming

Update Policy

Backward Compatibility

Best Practices

Model-Specific Optimization

For DeepSeek Chat:

For DeepSeek Coder:

For DeepSeek Math:

Performance Optimization

Error Handling

Migration Guide

From OpenAI Models

Migration Steps

Future Roadmap

Upcoming Models

Planned Features

Getting Help

Documentation

Support