By Paul Derrington, Jr.
CEO | Chief Engineer | Research Engineer | Editor, D-CSIL
3 min read
Key Takeaways
On November 13 DeepSeek opened the tools behind its chat app — search, files, link reading, safety checks and voice — to developers through its API. The built-ins run on its Flash model. Search and page fetches are billed at $0.75 per 1,000 each, on top of tokens.

What DeepSeek opened
DeepSeek says web search, file input, URL reading, voice interaction and safety checks work together with model inference in its app, and on November 13 it opened those capabilities to developers. Before the release, developers who wanted those features connected outside tools to the API and handled the hand-offs themselves, WebProNews reported, citing Chinese outlet The Paper.
There was one early step. On October 17, DeepSeek opened its in-house search engine as standalone Search and Fetch APIs. On November 13, DeepSeek added the other app features to the API and made search available as a built-in tool inside model calls.
Now: the app, as parameters
In its November 13 change log, DeepSeek says developers can now switch the app features on with parameters in the existing APIs, “with the tool calls and context assembly handled by the service.” The built-ins run on deepseek-flash — the API name for DeepSeek-V4.1-Flash-1017 — across the Messages, Responses and Chat Completions formats. DeepSeek's documentation lists the built-ins for deepseek-flash only, not for its V4 Pro model.
Turn on the deepseek_web_search tool with thinking disabled and the model runs web or image searches and answers with citations. Leave thinking on and the same tool becomes deep thinking search: the model searches and reads links on its own over multiple reasoning steps — up to about 20 tool calls before it answers, according to DeepSeek's web tools guide. A second tool, deepseek_read_url, fetches pages linked in a user's message. File input accepts PDFs, Office files and images, which are extracted or parsed and added to the conversation. A safety-check parameter reviews input and output and returns a refusal marker when it finds a risk.
For code that decides for itself when to call a tool, Search, Fetch, OCR, speech-to-text (ASR) and text-to-speech (TTS) are also sold as standalone APIs. DeepSeek also open-sourced a demo chat client built on the new APIs; it runs in the browser with no backend and can import a user's DeepSeek chat history.
What the pieces cost
The tools are billed on top of tokens, according to DeepSeek's Models & Pricing page. Search and Fetch each cost $0.75 per 1,000 successful requests. File parsing and OCR cost $0.225 per 1,000 pages. Speech is half price for a limited time — ASR at $0.058 per hour of audio (list $0.116) and TTS at $0.225 per 10,000 characters (list $0.45) — and the pricing page lists no end date for the discount. The model itself runs $0.30 per million uncached input tokens and $1.20 per million output tokens at peak hours, and half that off-peak.
On search, OpenAI's pricing page lists its web search tool at $10 per 1,000 calls and Anthropic's documentation lists $10 per 1,000 searches, each plus token costs: more than 13 times DeepSeek's $0.75. DeepSeek bills each query in a multi-query search call separately and does not bill failed or empty searches. In the other direction, Anthropic's web fetch tool carries no per-use fee beyond tokens, while DeepSeek charges for fetches at the same $0.75 rate as searches, so an agent that reads far more pages than it searches pays a charge on DeepSeek that it would not pay on Claude.
Conditions
DeepSeek says public-facing apps must enable the safety check, and that developers remain responsible for content moderation, applicable model registration and AI-generated content labeling. Its safety guide says the check does not replace a service provider's own content review.