Skip to content

DeepSeek exposes its app's search, OCR and voice tools via API

DeepSeek's 13 November release puts the whole DeepSeek App experience behind its API: built-in web search and file reading on deepseek-flash, plus OCR, speech recognition and text-to-speech as standalone endpoints.

Source: DeepSeek

DeepSeek has opened the features behind its consumer app to API developers for the first time. The 13 November release adds built-in web search, deep-thinking search, file input, URL reading and safety checks to its deepseek-flash model (V4.1-Flash), reachable through the Messages, Responses and Chat Completions APIs, with the tool calls and context assembly handled server-side.

Turning on search or link-fetching is a matter of passing a parameter, such as deepseek_web_search or deepseek_read_url, in an existing request. OCR, speech recognition (ASR) and text-to-speech (TTS) are also available as independent endpoints, so you can call them standalone or combine them, without wiring up your own search index, PDF parser or scraper first. DeepSeek has published an open-source frontend demo, including support for importing existing conversation history, alongside a build guide.

Why it matters: building a DeepSeek-app-style product used to mean assembling your own search, file-parsing and voice stack around the model. That is now one API surface, which lowers the cost of building an agent or chat product that searches, reads files and talks back.

The caveat: DeepSeek says tool calls cost more than plain tokens, but the announcement only shows a pricing breakdown image, not a rate in the text, so check the actual numbers before switching these on at any scale.

Share

Related reading

More from The Wire

All briefs