AI News

DeepSeek opens up its full suite API, V4.1-Flash supports web search and file parsing

Phoenix TechnologyUpdated Nov 14 10:45

AI summary

DeepSeek has integrated the App's capabilities—deep-thinking search, file upload, link reading, voice interaction, and text recognition—into its existing API, and the company says developers can replicate the complete DeepSeek App 1:1. V4.1-Flash now supports these capabilities under the Messages, Responses, and Chat Completions interfaces; after enabling parameters such as deepseek_web_search and deepseek_read_url, tool calling and context assembly are handled server-side. The company has also open-sourced a lightweight front-end demo that covers web search, file and image upload, read-aloud answers, voice input, and Markdown and Mermaid rendering.

Why read this

AI score 70

The original text lists the capabilities bundled into DeepSeek's new API, including web search, file parsing, and link reading, which can be used to assess how much in-house development work a self-built application requires.

Topics

Tags

Body · AI translation

DeepSeek's full suite API is officially open! You can build your own DeepSeek too. Xinzhiyuan, November 14, 2026, 07:47:56, from Beijing

Previously, the long list of features you found most convenient in the DeepSeek App and web version—including deep-thinking search, file upload, link reading, voice interaction, and even text recognition...

Now they're all packaged into the API!

You can easily implement complex conversation features in your own applications.

Officials even went so far as to say: developers can recreate the complete DeepSeek App 1:1!

"What do you mean, recreate the complete DeepSeek App 1:1??? This is the light of AI!"

"DeepSeek really does think about its users. It really does, I'm crying."

Previously, what did you have to go through if you wanted to recreate DeepSeek from scratch?

First, you had to know how to deploy a large model locally.

However, just integrating an API was far from enough—you also had to go to great lengths to connect external tools and handle call orchestration.

It was like a Frankenstein's monster.

But today, the full-suite API opened by DeepSeek directly flips the table for developers!

The various interactive features of the DeepSeek App are directly integrated into the original API. All the tedious tool calls and context stitching are handled entirely by the DS server.

Just enable the corresponding parameters, and you can summon the complete form of DeepSeek!

Enable the corresponding parameters in the original API, and the DeepSeek server automatically completes the relevant tool calls and context stitching.

Now, V4.1-Flash fully supports the following capabilities under the Messages, Responses, and Chat Completions interfaces.

In non-thinking mode, turn on the deepseek_web_search tool, and the model can automatically search web pages and images, with citations included.

After enabling thinking mode, it's even more incredible—the model can autonomously search the internet and read links during multi-step reasoning. The logic is maxed out!

File input

Just drop in PDFs, Office documents, or images, and the server automatically performs plain-text extraction or mixed text-image parsing, seamlessly stitching them into the conversation context.

Link reading

Turn on the deepseek_read_url tool, and when a user sends a link in the conversation, the model automatically fetches the web page and answers based on the context.

Built-in model safety

Enable the safety verification parameters to directly provide basic risk control, review input and output, and also support image review, making compliance one step away.

Others selling APIs are like selling water and electricity; DeepSeek is directly selling water plants and power stations!

DeepSeek's vision goes far beyond this

Even more, to make things convenient for developers, DeepSeek has split the underlying capabilities into independent interfaces that can be called separately or in combination!

From now on, DeepSeek can be used as an underlying tool library—call whichever one you need, so easy~

Based on this new version of the API, officials have also open-sourced a lightweight front-end demo.

It fully demonstrates features such as web search, deep thinking, file and image upload, read-aloud answers, voice input, and Markdown and Mermaid chart rendering...

Rendering responses: A single response from the Chatbot contains multiple content blocks. The above shows where each type of block corresponds in the DeepSeek web version.

It even supports directly importing historical conversations and seamlessly continuing the chat!

Moreover, the API Key, historical conversations, and images are all stored locally in the browser, maximizing the sense of security.

Developers who want to copy the homework can go straight for it.

This time, Big Fat Fish has truly "opened up its vision."

Generally speaking, when a company makes a useful feature, it wants to hide it away for fear of others learning from it.

But DS directly brought out its most treasured technology—not only providing the tools, but even feeding the open-source demo right to your mouth.

The official announcement ends like this: Actually, our DeepSeek App is just a basic version. Now we'll take care of the dirty work and hard labor. Stop reinventing the wheel—go build something big!

It frantically pushes technology forward itself, paving the way for everyone, letting developers stand on its shoulders to build the next hit.

Back then, the selfless sharing of underlying protocols made today's internet possible. Today, DeepSeek is reshaping the shared foundation of the AI era.

DeepSeek truly deserves to be called the light of humanity.

Source

Phoenix Technology