> ## Documentation Index
> Fetch the complete documentation index at: https://docs.zerotwo.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# DeepSeek

> DeepSeek models in ZeroTwo — specialized for coding, mathematics, and logical reasoning.

DeepSeek is a Chinese AI research company that has produced some of the most capable open-source models for coding and mathematical reasoning. DeepSeek's models are available in ZeroTwo for users who need strong performance on technical tasks.

***

## Available Models

| Model             | Type | Strengths                                                             | Plan Required  |
| ----------------- | ---- | --------------------------------------------------------------------- | -------------- |
| DeepSeek Chat     | Text | Strong general chat and instruction-following, efficient              | Pro+           |
| DeepSeek Reasoner | Text | Extended step-by-step reasoning for math, logic, and complex problems | Pro+ (Premium) |
| DeepSeek Coder    | Text | Specialized code generation, debugging, and technical problem-solving | Pro+           |

***

## Strengths

**Coding excellence:** DeepSeek Coder is purpose-built for code generation and has benchmarked highly against other coding-focused models. It handles a wide range of languages, frameworks, and complex programming tasks.

**Mathematical reasoning:** DeepSeek Reasoner uses an extended thinking approach (chain-of-thought reasoning) that makes it particularly effective for math problems, proofs, algorithms, and logic puzzles where showing work matters.

**Cost-effectiveness:** DeepSeek models offer strong performance at competitive efficiency, making them a good choice for technical users who need strong coding and math capability.

**Chinese language support:** As a Chinese AI lab, DeepSeek models have strong Chinese language understanding alongside their technical strengths.

***

## Extended Thinking (Reasoner)

DeepSeek Reasoner supports **extended thinking** — a mode where the model works through problems step-by-step before producing an answer. This is particularly effective for:

* Complex multi-step math problems
* Formal logic and proofs
* Algorithm design
* Problems where the answer process matters as much as the result

When using DeepSeek Reasoner, you may notice longer response times for complex problems — this is expected as the model works through its reasoning chain.

***

## Best Use Cases

<CardGroup cols={2}>
  <Card title="Complex coding tasks" icon="cpu">
    Code generation, debugging, refactoring, code review, and architecture design across many languages.
  </Card>

  <Card title="Mathematics" icon="book">
    Algebra, calculus, statistics, proofs, and numerical problem-solving. DeepSeek Reasoner excels here.
  </Card>

  <Card title="Logic and algorithms" icon="list">
    Algorithm design, logical reasoning, puzzles, and step-by-step problem decomposition.
  </Card>

  <Card title="Technical analysis" icon="layers">
    Data analysis, scientific computing, and quantitative tasks that blend code and math.
  </Card>
</CardGroup>

***

## Choosing a DeepSeek Model

| If you need...                             | Use...            |
| ------------------------------------------ | ----------------- |
| Best coding performance                    | DeepSeek Coder    |
| Complex math or logic with reasoning steps | DeepSeek Reasoner |
| General chat with technical strength       | DeepSeek Chat     |

***

<Note>
  DeepSeek models are developed by a Chinese AI research company. For users with data residency requirements, review the [Data & Compliance](/privacy/data-and-compliance) page for information on how ZeroTwo handles API calls to different providers.
</Note>