DeepSeek's API now includes the features behind its own chatbot: web search with citations, reading links and files, a content safety check, automatic chat titles, and voice. Each is a switch you turn on with a request parameter, so you can build a ChatGPT-style assistant on one API key instead of stitching together several vendors.
DeepSeek App API at a glance
Search and Fetch | $0.75 per 1,000 successful requests each |
|---|---|
Model | deepseek-flash |
Token price | From $0.15/1M input (cache miss, off-peak) and $0.60/1M output |
Endpoints | Anthropic-style Messages, Responses, Chat Completions |
Files | PDF, Word, PowerPoint and images |
What's included
Web search: the model searches the web or images and answers with citations. With thinking on, it can search and read pages across several reasoning steps.
URL reading: links users paste are fetched and used as context.
File input: PDF, Word, PowerPoint and image files, parsed into the conversation.
Safety check: reviews inputs and outputs and refuses when it detects risk.
Titles and user context: automatic conversation titles and the user's date, location and language passed to the model.
Voice: speech-to-text and text-to-speech, plus OCR, as separate APIs.
Pricing
Search and Fetch each cost $0.75 per 1,000 successful requests, charged on top of tokens. deepseek-flash costs $0.15 per million input tokens on a cache miss and $0.60 per million output tokens off-peak, and double that during peak hours (01:00-04:00 and 06:00-10:00 UTC on weekdays). Cached input is far cheaper. The pricing page doesn't list voice or OCR prices.
What to know
With web tools, a high safety setting or titles turned on, DeepSeek overrides parameters like temperature, top_p and stop.
The output limit applies to each reasoning step, not the whole reply.
The safety check is basic. If you serve the public, compliance (content review, registration, AI labeling) is on you.
Your users' queries and files go through DeepSeek's platform, so read the terms before shipping.
Our take
Search is the add-on that quietly blows up AI app budgets, and $0.75 per 1,000 is low enough to leave it on by default. Bundling files, OCR and voice under one key removes several vendor contracts from a typical build. The trade-off is data residency and compliance, which matter more than price for many teams.
DeepSeek App API alternatives
Gemini Decisions API: cheap, fast yes/no and multiple-choice calls for agents.
ElevenLabs Scribe v3: dedicated speech-to-text with live speaker labels.
Frequently asked questions
Can I use OpenAI or Anthropic SDKs?
Yes. DeepSeek offers Messages, Responses and Chat Completions endpoints, so existing SDK code mostly needs a new base URL.
Is there a demo?
Yes, an open-source chat demo. DeepSeek says it's for demonstration only and stores your API key in the browser.
As featured in World of AI
DeepSeek puts its whole app into the API (Nov 13, 2026)
Facts checked against DeepSeek's build guide and pricing page on November 13, 2026.
Want four tools like this every weeknight? Subscribe to World of AI for free. Browse more in our AI Tools directory.