Text models · Developers

DeepSeek opens its app features as APIs: web search, file input, URL reading and safety checks

DeepSeek's API now offers web search, file input, URL reading and a safety check on deepseek-flash, plus standalone OCR, speech-to-text and text-to-speech.

DeepSeek posted “Releasing the full suite of application APIs” in its API change log on November 13, 2026. The features that power the DeepSeek app (web search, file reading, link reading and safety checks) are now parameters you can switch on in the API, with DeepSeek’s service handling the tool calls and context assembly.

What to know

  • Model and APIs: the features work with deepseek-flash in the Messages (Anthropic-style), Responses and Chat Completions APIs, according to DeepSeek’s “Build a Chatbot” guide.
  • Web search: add the deepseek_web_search tool. With thinking off, DeepSeek says the model searches the web or images when needed and cites its sources. With thinking on, it can search and read links over several reasoning steps.
  • URL reading: add the deepseek_read_url tool and the model fetches links found in the user’s message and uses the pages in its answer.
  • File input: DeepSeek says PDFs, Office files and images are parsed and added to the model’s context. Its guide lists PDF, Word, PowerPoint and image files.
  • Safety check: a safety_check parameter (the guide’s example uses {"level": "high"}) reviews inputs and outputs and returns a refusal marker when it detects risk. DeepSeek says public-facing services should use it and meet their content review, model registration and AI labelling obligations.
  • Standalone APIs: Search and Fetch, OCR, speech-to-text (ASR) and text-to-speech (TTS) are also available on their own. DeepSeek also published an open-source demo chat client that can import DeepSeek chat history.

Prices

DeepSeek’s Models and Pricing page lists $0.75 per 1,000 web searches and $0.75 per 1,000 URL fetches, $0.225 per 1,000 pages for file parsing and OCR, and text-to-speech at $0.45 per 10,000 characters and speech-to-text at $0.116 per hour, both with a limited-time 50% discount. No price is listed for the safety check. deepseek-flash token prices also vary between off-peak and peak hours, so check the pricing page before you estimate costs.

Why it matters

Until now, building a chatbot that searches the web and reads uploaded files meant wiring up your own search provider, scraper and document parser. DeepSeek now offers all of it behind one model call, at low per-call prices. The trade-off is less control over which sources are used, so tell the model how to cite and when not to search.

Our own system prompts for this kind of assistant are in how to write prompts for an AI chatbot with web search. Background: the October Search and Fetch APIs, the DeepSeek page and the model tracker.

Source: DeepSeek API docs. We summarise in our own words; check the source for the full details.

Related

More AI news

stay in the loop

New guides, prompts and AI model news by email