DeepSeek Opens Up Its App's Features Through the API
DeepSeek has opened up its app's features through the API: web search, file parsing, and speech recognition, via the deepseek-flash model.
DeepSeek, a Chinese AI lab, moved features from its own app into its API, the channel other programs use to call its models. The change runs through a model called DeepSeek-Flash. The main addition is web search with cited sources. Alongside it come step-by-step reasoning search, parsing of PDF and office files, and speech recognition and synthesis.
Before this, teams building products on DeepSeek had to bring in outside services for search and speech recognition. Now these turn on with a few extra settings inside the same API call, next to the standard formats programmers already use to talk to the model: Messages, Responses, and Chat Completions. That removes one integration step for anyone building on DeepSeek.
The extended features cost more than plain text processing; DeepSeek's announcement does not give exact prices, pointing instead to a separate pricing page. It is not known whether the quality of this search and file parsing matches the built-in tools from OpenAI and Anthropic, the companies behind ChatGPT and Claude.
Source: api-docs.deepseek.com
Free course
Stop reading about AI — start building with it
The free Claude Code course: your first site, tool or game — no coding. No upsells, no cross-sells — nothing to buy here.
Start free →
Author
Evgenii Arsentev
PhD · Founder & CEO, Dad4Kids
Articles · Latest articles