> For the complete documentation index, see [llms.txt](https://docs.cosverse.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.cosverse.ai/ai-models/deepseek.md).

# DeepSeek

DeepSeek AI models in Cosverse AI help with AI chat, coding, reasoning, research, summarization, and document analysis.

Use DeepSeek models when you need powerful reasoning, fast responses, and high-quality code generation across a wide range of tasks.

#### What you can do with DeepSeek models

* Ask questions and get clear, detailed AI chat responses.
* Write, debug, and explain code across multiple programming languages.
* Solve complex reasoning and math problems step by step.
* Summarize long documents and extract key information.
* Brainstorm ideas, research topics, and organize information.
* Analyze files and work from detailed context.

#### Available DeepSeek Models:

<table><thead><tr><th width="219.65234375">Model Name</th><th>Short Description</th></tr></thead><tbody><tr><td>DeepSeek V3.2</td><td>A large language model with strong reasoning, coding, and instruction-following capabilities.</td></tr><tr><td>DeepSeek V3 (US-Hosted)</td><td>A lightweight DeepSeek model optimized for fast and reliable responses, hosted in the US.</td></tr><tr><td>DeepSeek R1 (US-Hosted)</td><td>A fast, lightweight reasoning model for logic and math tasks, hosted in the US.</td></tr><tr><td>DeepSeek V3.2 Speciale</td><td>A high-compute variant of DeepSeek V3.2 for demanding and complex generation tasks.</td></tr><tr><td>DeepSeek V3.1</td><td>A lightweight hybrid reasoning model that balances speed and intelligence for everyday tasks.</td></tr></tbody></table>

#### Why use DeepSeek Models in Cosverse AI

DeepSeek models are developed by DeepSeek AI and are known for strong coding, reasoning, and multilingual capabilities. Here's why they stand out in Cosverse AI:

Strong coding and reasoning — DeepSeek models excel at writing, debugging, and explaining code across many programming languages.

Fast US-hosted options — DeepSeek V3 and R1 US-Hosted variants deliver low-latency responses for faster workflows.

High-performance thinking — R1 and V3.2 Speciale are built for deep reasoning and complex multi-step problem solving.

#### Best use cases for DeepSeek Models

* Writing, debugging, and reviewing code in any programming language
* Solving complex math, logic, and reasoning problems step by step
* Generating technical documentation and explanations
* Summarizing and analyzing long documents or research papers
* Brainstorming ideas and structured content creation
* Fast AI chat with low latency using US-hosted models

#### How to get better results with DeepSeek Models

Be specific and structured — Clear, well-defined prompts give DeepSeek the best context to respond accurately.

Use R1 for reasoning tasks — Choose DeepSeek R1 when you need step-by-step thinking for math or logic problems.

Use V3.2 Speciale for demanding tasks — For long, complex generation or high-quality output, V3.2 Speciale performs best.

Use US-Hosted for speed — DeepSeek V3 and R1 US-Hosted models are ideal when response time matters.

Provide code context — When debugging or reviewing code, include the full code block and describe the expected behavior.

If you are comparing providers, see [Available AI Models](/getting-started/available-ai-models.md).