DeepSeek
DeepSeek: free AI chat, thinking and search
Use DeepSeek on the web or mobile for AI conversation with thinking and search options. Developers can access its models through a separate token-priced API.
FreeAccount required

- Access
- Free
- Usage pricing
- USD 0.15 / 1,000,000 input tokensUncached input — off-peak
- Available on
- Web, Mobile, API
Free AI conversation, with a paid API for developers
DeepSeek is a free AI assistant on the web, iOS and Android. It offers conversation with thinking and search options, making it an accessible starting point for questions, writing and research. The hosted chat service requires an account, and the mobile app keeps conversation history alongside the chat interface.
Developers have a separate route through DeepSeek’s paid API. DeepSeek-V4.1-Flash is the current Flash model, requested as deepseek-flash, and supports visual inputs alongside text. The API also offers thinking and non-thinking modes, tool calls and structured JSON output for applications built around the model.
API costs depend on how much text and image content goes into a request and how much output comes back. Cached input, uncached input and output have different rates, with lower prices outside the weekday peak windows. That makes the complete request cost more useful than a single headline input price. The free chat app does not use these per-token charges.
DeepSeek Harness extends the range into desktop tasks. Available in public preview, it can work with local files, documents, data and code, with plugins adding further tools. Windows and Apple Silicon Mac downloads are available. Harness is a separate companion; its task and file features are not the same as the mobile chat app.
DeepSeek also publishes open model releases for developers who want other deployment options. Independent hosting requires hardware, deployment and operations to manage. DeepSeek’s hosted service follows its own privacy policy, which states that personal data is processed and stored in China and describes training opt-out rights.
Best for
- People looking for free AI chat on the web and mobile
- Developers building with a token-priced text and vision API
- Technical teams exploring open models and desktop agent tools
Limitations
The hosted assistant uses cloud processing. DeepSeek says it processes and stores personal data in China; running an open model independently has different data arrangements.
DeepSeek features
Free web and mobile chat
Use DeepSeek for AI conversation on the web, iOS and Android. The mobile app includes conversation history, and the hosted chat service requires sign-in.
Thinking and search
Thinking and search options support longer questions and current-information research. DeepSeek’s search service can share query keywords with third-party search providers.
Visual analysis through the API
The current Flash API accepts JPEG, PNG, GIF and WebP images alongside text. It can describe pictures, read screenshots and analyse charts. These are developer API capabilities; attachment handling in the chat app is separate.
Developer tools and structured output
The paid API supports tool calls, JSON output, and OpenAI- and Anthropic-compatible interfaces. Flash requests can use a 1M-token context, with account-level concurrency limits.
Open model releases
DeepSeek publishes model releases and technical material for developers exploring deployment options. Independent hosting gives teams a separate route, with hardware and operations to manage.
DeepSeek Harness desktop preview
Harness can work on documents, spreadsheets and code, including longer tasks involving local files. Plugins extend its tools. The desktop preview is available for Windows and Apple Silicon Macs running macOS 13 or later.
DeepSeek pricing
The web and mobile assistant is free. Developer API calls are paid by usage. The examples below show V4.1-Flash rates per one million tokens for uncached input, cached input and output during off-peak hours. Peak rates are twice these amounts. Peak windows are 01:00–04:00 and 06:00–10:00 UTC on weekdays, excluding Chinese public holidays; all other hours are off-peak.
Example plans and usage rates; additional charges may apply.
Uncached input — off-peak
USD 0.15 / 1,000,000 input tokens
DeepSeek-V4.1-Flash API
Plan details
Per one million tokens, outside peak windows. Peak: 01:00–04:00 and 06:00–10:00 UTC on weekdays, excluding Chinese public holidays. API balance billing; separate from free chat.
Cached input — off-peak
USD 0.003 / 1,000,000 input tokens
DeepSeek-V4.1-Flash API
Plan details
Cache-hit input only. Per one million tokens, outside peak windows. Peak: 01:00–04:00 and 06:00–10:00 UTC on weekdays, excluding Chinese public holidays. API balance billing; separate from free chat.
Output — off-peak
USD 0.6 / 1,000,000 output tokens
DeepSeek-V4.1-Flash API
Plan details
Per one million tokens, outside peak windows. Peak: 01:00–04:00 and 06:00–10:00 UTC on weekdays, excluding Chinese public holidays. API balance billing; separate from free chat.
Technical specifications
AI assistants
| Feature | DeepSeek |
|---|---|
| File inputs | JPEG, PNG, GIF and WebP image inputs for visual analysis.DeepSeek-V4.1-Flash developer API |
| Web search and citations | The consumer interface offers Search. Search queries can be shared with third-party search providers.Hosted consumer search |
| Usage limits | 1M-token context for API requests.DeepSeek-V4.1-Flash developer API |
| API access | SupportedPaid hosted developer API |
DeepSeek alternatives
ChatGPT combines conversation with files and creative tools, Claude emphasises writing and document work, and Gemini connects with supported Google apps. Their reviews below cover the features, prices and limits of each assistant.
DeepSeek FAQs
Is DeepSeek free?
Yes. The official web and mobile assistant is free. Developer API calls are billed separately for input and output tokens, with discounted rates for cached input and off-peak use.
Which model name should developers use?
Use deepseek-flash for the current DeepSeek-V4.1-Flash API. The older deepseek-v4-flash and deepseek-v4-flash-vision-exp names remain accepted for compatibility, with requests served by the current Flash model and billed at Flash rates.
Are DeepSeek’s API limits also its chat limits?
No. The Flash API’s 1M-token context and its image-request limits apply to developer requests. The free chat app has its own attachment handling and usage limits; a paid API specification does not extend those limits in the app.
Where is hosted DeepSeek data stored?
DeepSeek’s privacy policy states that it directly collects, processes and stores personal data in China. It also describes the right to opt out of using personal data for model training. Independently hosted model deployments have their own data arrangements.
How much does the DeepSeek API cost?
For V4.1-Flash during off-peak hours, one million uncached input tokens cost USD 0.15, cached input tokens cost USD 0.003, and output tokens cost USD 0.60. Peak rates are double. Both the input and output contribute to the bill; the cached rate applies only to cache hits.
Can DeepSeek understand images?
The current deepseek-flash developer API supports visual input in JPEG, PNG, GIF and WebP formats. You can send images with text to analyse screenshots, charts and other pictures. Its image-request limits are separate from consumer chat uploads.
What is DeepSeek Harness?
Harness is DeepSeek’s separate desktop agent companion, currently in public preview. It can work with documents, data, code and local files, with plugins extending its tools. Downloads are available for Windows and Apple Silicon Macs running macOS 13 or later.
Is this your product? Claim this page to update your listing.