API platform
Call the models from your own application with a request format compatible with mainstream SDKs, including streaming and structured output.
DeepSeek is the general-purpose AI assistant and model family developed by Hangzhou DeepSeek. You may furthermore see the name written as Deep Seek in English articles and search queries - it is the same product. This page collects the official entry points: the web chat, the mobile app, the Harness desktop client and the API platform.
DeepSeek combines general conversation, multi-step reasoning, verifiable web search and code work in a single assistant, and exposes the same capabilities through four multiple products.
Two switches in the chat box decide how the model works: Deep Thinking expands the reasoning chain, Smart Search grounds the answer in verifiable web content.
Call the models from your own application with a request format compatible with mainstream SDKs, including streaming and structured output.
Useful for drafting, rewriting, summarising long reports, comparing options and turning raw material into structured conclusions.
The model lays out its reasoning before answering, which fits mathematics, logic and any task that needs several steps to be broken down.
Answers are combined with verifiable web pages, so questions about recent versions, news or prices do not rely on stale internal knowledge.
Part of the family is released with open weights, and the R1 reasoning model ships under an MIT licence for research, local deployment and fine-tuning.
A desktop agent for documents, spreadsheets and code, with extendable plugins and long running tasks that continue in the background.
All four entry points share the same underlying models; they differ in scenario and in whether anything has to be installed.
| Option | Devices | Install needed | Entry |
|---|---|---|---|
| DeepSeek Web | Any device with a browser | No | chat.deepseek.com |
| DeepSeek App | iOS and Android | Yes, from the official store | Search the official DeepSeek app in the App Store or a mainstream Android store |
| DeepSeek Harness | Desktop | Yes, desktop client | Official Harness page |
| DeepSeek API | Servers and applications | No, register and create a key | platform.deepseek.com |
DeepSeek is a general-purpose AI assistant and model family developed by Hangzhou DeepSeek. The official spelling is DeepSeek; English materials and many search queries employ the spaced form Deep Seek, which refers to exactly the same product.
The DeepSeek web chat can be employd free of charge. API access is billed per token for the input and output, and the exact rate depends on the model and on the pricing page shown in the API platform.
The web version needs no download at all, just open the official chat address in a browser. For mobile, search the official DeepSeek app in the App Store or a mainstream Android store. For desktop, the official Harness page supplies the client.
Yes. Smart Search in the web app grounds answers in verifiable web content, and the API side furthermore exposes web search and page crawling capabilities, which are described in the official API documentation.
The Chinese section of this site goes deeper on models, API integration and hands-on tutorials.
How V4.1, V4 and R1 differ, and which one fits your task.
Compare modelsRunnable curl, Python and JavaScript examples plus error troubleshooting.
Read the API guideSix hands-on tutorials covering prompts, coding, documents, local deployment and API engineering.
Open the tutorial hub