DeepSeek App API: search, files, OCR and voice. Price the tools
The DeepSeek app API now includes web search, file input, URL reading, OCR, ASR and TTS on deepseek-flash. Tool fees are extra, and public apps must use safety checks.
The DeepSeek app API release puts the full consumer app behind one key. On November 13, DeepSeek published "The full DeepSeek App experience, now via API": web search, file input, URL reading, safety checks and voice, all on the low-cost deepseek-flash model. A small team can now build a chatbot that reads PDFs and cites the web without five vendors. But each tool has its own meter, and one rule changes how you deploy.
What actually happened
From DeepSeek's announcement and its pricing page:
- Built-in tools. Web and image search with citations, a deep thinking search mode that reads pages during reasoning, file input for PDFs, Office files and images, and automatic URL reading.
- Standalone APIs. Search, Fetch, OCR, ASR (speech to text) and TTS (text to speech).
- One model, three formats. The tools run on
deepseek-flash(V4.1-Flash) through the Chat Completions, Responses and Anthropic-style Messages formats. - The prices. Search and fetch: $0.75 per 1,000. OCR and file parsing: $0.225 per 1,000 pages. ASR: $0.116 per audio hour. TTS: $0.45 per 10,000 characters. ASR and TTS are at half price for a limited time.
- Tool fees are extra. DeepSeek says the tools "incur tool usage fees in addition to standard token charges." Safety review tokens are billed as output tokens.
- A demo you can read. The chat-demo repository is MIT-licensed.
What the DeepSeek app API rules mean for your business
The numbers are low. OCR at $0.225 per 1,000 pages makes invoice and receipt extraction almost free at small-shop volume. But read three lines before you build.
Safety checks are not optional. The announcement says "Public-facing apps must enable safety checks." It also says the developer stays responsible for content moderation, model registration and AI content labels. That is a compliance duty, not a feature.
You lose some controls. The docs say that when search, URL reading or high-level safety checks are on, the service ignores parameters such as temperature and top_p. If your prompts depend on a low temperature, test again.
The data path is not described. The pages we read do not say where uploaded files, audio or search queries are processed or how long they are kept. An invoice is customer data. A support call recording is customer data.
How we would use it:
- Count calls, not only tokens. One deep search can fire many searches and fetches. Log tool calls per task before you set a budget.
- Route by sensitivity. Public documents and product specs can go to the cheap path. Keep customer files on a provider whose data terms you have read.
- Keep an adapter. OCR, speech and search are commodities now. One interface, many providers.
- Budget for list price. The ASR and TTS discount has no end date in the docs.
Key takeaways
- DeepSeek opened its app features to the API on November 13: search, file input, URL reading, OCR, ASR and TTS
- OCR and file parsing cost $0.225 per 1,000 pages; search and fetch cost $0.75 per 1,000
- Tool fees come on top of token charges, and safety review tokens bill as output
- Public-facing apps must enable safety checks, and the developer keeps the moderation duty
- The docs do not describe data location or retention for files, audio or queries
Do you want cheap OCR and search without a single-vendor dependency? We build document and agent pipelines with provider-agnostic adapters, routing by data sensitivity and per-task spend caps. See our AI automation work or estimate the return first.
Sources: DeepSeek: The full DeepSeek App experience, now via API, DeepSeek API docs: Models & Pricing, deepseek-ai/chat-demo on GitHub.
- #deepseek
- #api-pricing
- #ocr
- #web-search
- #ai-agents
Tommy Rush — Founder, Rush Commerce
Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More
Get The Rush Report weekly — one email, zero fluff.
Keep reading
Neo4j Relate: an embedded graph database for AI agents
Neo4j Relate is an open-source embedded graph database for AI agents, written in Rust with MCP included. Test it for agent memory, and read the install script first.
Read itSalesforce SOV-Transcribe-1: open-weight speech-to-text, Apache 2.0
Salesforce SOV-Transcribe-1 is an open-weight speech-to-text model under Apache 2.0 with a claimed 2.28% WER. It's batch only. Test it on your own call recordings.
Read it