As soon as you open your eyes, WeChat confirms:

Formally connectedDeepSeek-R1orlit. full-blooded edition!

The path of use is very simple: open WeChat - click on the search box - select the middle of the page "AI search" - select the deep thinking mode - -Use it~

The generated results alsoWith social attributes, you can share and forward to friends, or forward to the circle of friends.

Many people have previously called for wanting to be able to share DeepSeek's answers with a single click, but the long-cherished wish has not been realized. Before last night, almost all the ways to share the results generated by DeepSeek were screenshots (and since it would show the reasoning process, the screenshots to be taken were extraordinarily long).

Now, this need is finally being implemented by Microsoft Real.

Public number "consensus crusher" overnight calculations, the current WeChat can support 50-100 million users online users (online at the same time, not DAU).

This number, which basically satisfies the dosage of WeChat's first Deepseek users, alsoExceeds ChatGPT's concurrent users.

Although it's still in gray scale, this is clearly a move by Tencent to continue to fully embrace DeepSeek and really get right up and running with it.

Good news, Quantum Bit Edit was then the first to be grayed out.

Joy to the world, this shall not be tested in real life? --

Test WeChat access to DeepSeek

WeChat's AI search currently offers two modes:

• Quick Answer.: Generate profile answers, suitable for generalized simple questions.
• reflect in depth: DeepSeek-R1 as a modeling support, long thinking, full answers, more comprehensive and detailed answers.

So to experience R1's in depth, you need to select 'Deep Thinking Mode'.

First, let's see if it supports basic networking modes.

In the past two days, Yushu robots and the whole flower work, they updated the dance video, more flexible than the Spring Festival Gala and more stunning, and even stunned the crooked people.

Throw this question to WeChat:Ushu Robot's latest dancing video, what is the dance.

The answer received was:

It responds without much lag and outputs quickly. The first thing it showed was its reasoning process, after an internet search it was determined that this was theFebruary 14, 2025progress, and then provided additional information considering the user's interest in robotics, culminating in the following answer.

There's a small bug here in that there's no clear distinction between the thought process and the formal answer, and the typography doesn't come out. But I guess it's because the grayscale test wasn't very stable, and there were no problems after that.

From the 10 references it attaches, the first 7 are derived from thepublic number. This also proves that there is a huge content ecosystem behind this AI search to support WeChat.

Okay, so the basic networking reasoning function is there, so let's get some difficulty on it.

Previously, people used DeepSeek to use the whole job, but there are fortune-telling, persona creative writing and so on.

As it is:

Write a paragraph on the theme of WeChat's access to DeepSeek in the tone of Xin Qiji.

At the end of the reflection, it hides the thought process to give a formal answer, which can also be expanded at any time by clicking on it.

The final answer given is as follows:

Along with the creation of the diagrams, notes are given, which conceal a number of mysteries.

The first half of the paragraph represents the significance of the event, "Zhi Kai Xing Han" represents a double meaning, and the second half reveals functional attributes, such as tuning into WeChat's ecological data.

After answering the question, it will suggest a related question or "ask another question".

In this way, it is possible to continue answering all the way through by clicking on a question or continuing to ask it (but continuous answering in conjunction with the above is not supported).

And as long as you don't skip out, all the quizzes are presented in the same interface.

Overall use, because in the gray-scale testing and the limitations of the experience window (mobile support only), there are some areas that need to be improved.

For example, search is only supported on mobile, and information shared to friends and circle of friends cannot be viewed on computer.

In addition to this, the answers are not as diverse as the official one, which should also be limited by the output token and the inability to do what the official one does - being able to find historical quizzes as well as contextualized consecutive answers of this kind.

In this way WeChat's access to DeepSeek presents still a kind ofshort, plain, and fastThe search method of WeChat is also quite in line with the attribute of instantaneous and private socialization.

While the world is still fretting over AI portals ......

What are the effects of WeChat's access to DeepSeek and what are the emergent effects?

A word of explanation:

When the global technology giants are still scrambling for AI entrances, Tencent used a seemingly light "ecological graft", which was jokingly called "the end of the AI search competition" by netizens.

Does WeChat's move live up to that praise? Let's sort it out.

First of all, the meat measurement of physical sensation, the same access to DeepSeek-R1, used in Tencent Yuanbao, is faster than in the micro-credit, less card.

But it doesn't matter.

Importantly WeChat's access to DeepSeek-R1 is a zero-learning-cost entry setup in the face of the unquantifiable mass of DeepSeek users.

In this way, 1.67 billion WeChat users(Data from internet sources, as of October 2024)No need to download a new app, no need to register a new account, you can get direct access to the strongest current inference big models in the Chinese language field.

Secondly, and the real killer app, lies in the fact that this move has integrated the data resources of R1 and WeChat ecosystem with the depth of RAG (Retrieval Augmented Generation) technology under the premise of providing a simple and easy-to-reach portal.

When users ask "how to book tickets for the Forbidden City", DeepSeek can not only search the whole network, the most special thing, it can be in the WeChat public number, video number, small program to capture the relevant information, can catch the call of the official small program of the Forbidden City service interface ---. This is really an exclusive ecological resource of Tencent.

The clear feedback at the end with emoji added answers makes it easy to read to catch the point, and it's kinda human.

And as we've previously noted, Tencent's embrace of DeepSeek has been broader and more comprehensive.

Previously, there have been cloud platform Tencent Cloud, Tencent Cloud's large model knowledge application development platform knowledge engine, AI intelligent workbench ima, the main AI application Yuanbao and other Tencent family, has access to DeepSeek-R1.

Open source is so winning, Tencent is so open, what is the ultimate result orientation?

It's the user who can put good AI in their phone and use it with the lowest learning cost, the easiest way, and the widest range of uses.

In its latest tweet, the public website Consensus Crusher called the move "a turning point for 2C apps".

There's also some information revealed inside:

△Captured from the tweet "WeChat + Deepseek: A turning point for 2C apps" by the public number "Consensus Crusher".

The super apps of the mobile internet era are still being utilized today to take advantage of the super apps, along with the brains that created them.

All in all, starting today, the most commonly used application by Chinese people + the strongest inference model in China + ecological data resources of very good quality are so combined together by WeChat and DeepSeek Water Spirit.

One More Thing

Yesterday Perplexity launched Deep Research for free.

In terms of performance, Humanity's Last Exam, which has been a big hit lately in testing AI's reasoning ability, was used and achieved an accuracy of 20.51 TP4T.

Outperforms Gemini Thinking, o3-mini, o1, DeepSeek-R1 and many other mainstream models.

In response, Perplexity's CEO publicly thanked DeepSeek:

That's because DeepSeek is open source, cheap and fast.

But Altman couldn't sit still - this morning, he shouted out on Twitter, taking aim at the Perplexity CEO.

He said:

ChatGPT (4o) is the best search product on the web!

But Now?

DeepSeek didn't say anything, and Weebly smiled without saying a word.

Reference Links:
[1] https://x.com/sama/status/1890816782836904000
[2] https://mp.weixin.qq.com/s/V8Hri4WSVJWgp8IcKxlIvw

This article is from WeChat:Quantum Bit (ID: QbitAI)Author: Bai Jiao, Hengyu