DeepSeek opens the tools behind its chat app to developers through its API
DeepSeek now lets developers switch on web search, file input, URL reading and safety checks in its API, and sells OCR, speech recognition and text-to-speech as separate endpoints.
DeepSeek announced on November 13, 2026 that the features people use in its consumer app are now part of its developer API. In a post on its API documentation site, the company says developers can turn these features on with a few parameters, or call some of them as standalone services, and even rebuild a full copy of the DeepSeek app inside their own products.
What's new
- Built-in tools: according to DeepSeek, its V4.1-Flash model (named
deepseek-flashin the API) now supports web search, a search mode that keeps reading pages while the model reasons, file input for PDFs, Office documents and images, automatic reading of links in user messages, and safety screening. DeepSeek says the tool calls and context assembly run on its servers. - API formats: the company lists support across its Messages, Responses and Chat Completions interfaces.
- Standalone services: Search, Fetch, OCR, speech-to-text (ASR) and text-to-speech (TTS) can also be called on their own. China's QbitAI notes that voice is the first time DeepSeek has offered official speech endpoints.
- Safety rule: DeepSeek says apps open to the public must enable the safety checks, and that developers stay responsible for content moderation and labelling of AI-generated content.
- Demo: the company released an open-source chat front end on GitHub that can import DeepSeek conversation history.
Availability and pricing
Tool use is billed on top of normal token charges. DeepSeek's pricing page lists $0.75 per 1,000 searches or web fetches and $0.225 per 1,000 pages of file parsing or OCR. Speech is on a limited-time half-price offer: TTS at $0.225 per 10,000 characters (list price $0.45) and ASR at $0.058 per hour of audio (list price $0.116).
Why it matters
Analysis: developers who wanted search, document reading or voice on top of DeepSeek previously had to wire up separate services and glue them to the model themselves. Handling that on DeepSeek's side lowers the effort of building assistants, note-takers and document tools on its models, and the low listed speech prices put pressure on other voice API vendors.
Source: DeepSeek