In a significant move, WeChat has begun integrating DeepSeek, expanding its network and capabilities in the burgeoning field of Artificial Intelligence (AI).

Recently, WeChat users discovered a new feature in the search bar labeled "AI Search." When selecting this option, users are presented with two choices: "Quick Answer" or "Deep Thinking." The "Deep Thinking" mode utilizes the DeepSeek-R1 model, allowing for more thoughtful and comprehensive responses to inquiries.

On February 16, 2023, Tencent announced that WeChat had officially begun a grayscale test of DeepSeek integrationUsers selected for this test can see the "AI Search" option in the search input at the top of the dialogue box, granting them free access to the full version of the DeepSeek-R1 model.

This integration has sent ripples through the AI industry, indicating a shift in the competitive landscape.

Tencent's Full Embrace of DeepSeek

In early February, Tencent Cloud announced the launch of the DeepSeek-R1 and V3 model APIsThe Knowledge Engine, Tencent Cloud's platform for developing knowledge applications, also incorporated these models and now supports online search, making it one of the first cloud vendors to integrate DeepSeek API with online search capabilities.

Just two weeks later, Tencent escalated its strategy by announcing the integration of DeepSeek into WeChat, signifying a deep commitment to this model.

Users voiced concerns over the partnership, questioning whether DeepSeek would access their private Moments or public accounts

Advertisements

Tencent addressed these worries directly.

It is well known that the content on WeChat's platform is not fully shared with other search engines, remaining a part of Tencent's "private domain" trafficTherefore, integrating DeepSeek implies that Tencent will "share" the content available on WeChat with DeepSeekUsers can benefit from networked searches, tapping into rich content from official accounts and the broader internet, provisioned by DeepSeek, to enhance the quality of responses.

Although this feature has been rolled out, Tencent noted that it remains in a grayscale testing phase, refining its offerings based on user feedback and experiences.

Why was the large model integrated into WeChat's search function? Tencent's reasoning lies in the improvement of search intelligence and accuracy; this includes better comprehension of user intent and the capacity to process complex queriesBy adopting models like Hidden Yuan and DeepSeek, Tencent is enriching the user's search experience.

A primary concern among users is whether AI Search will utilize their private chat records or MomentsTencent confirmed that AI Search only consolidates information from public accounts and other publicly available information on the internet, ensuring that personal data remains untouched.

Tencent is also exploring deeper integration of DeepSeek across its various products, such as the Tencent Cloud AI Code Assistant and Tencent Yuanbao, to enhance user engagement and service offerings.

On February 13, Tencent's AI assistant "Tencent Yuanbao" announced that it would support both the Hidden Yuan and DeepSeek models, leveraging the full capabilities of DeepSeek-R1671B alongside connected search, integrating information from WeChat's official accounts and video accounts.

By February 16, the QQ Browser had also officially integrated the full version of the DeepSeek-R1 model, supporting deep thinking, online search, multi-round dialogue, and history recall features.

By adopting DeepSeek-R1, a self-developed Hidden Yuan model, Tencent is striving to enhance its AI search capabilities within the WeChat ecosystem while avoiding the substantial investments often required for self-developed models

Advertisements

This integration allows for customized responses based on stored chat records and official account content, marking a significant step toward intelligent customer service expansion.

The Rise of the DeepSeek "Friend Circle"

DeepSeek's emergence has profoundly impacted the industry landscape. Since February, various platforms including Baidu Smart Cloud, Huawei Cloud, Alibaba Cloud, Tencent Cloud, 360 Digital Security, and Cloud Axis Technology have announced the launch of DeepSeek models, enabling users to access DeepSeek-R1, DeepSeek-V3, and other models across these platforms.

On February 1, Huawei Cloud reported a collaboration with Silicon-based Fluid to release DeepSeek R1/V3 based on Huawei Cloud's Ascend Cloud service.

On February 3, Baidu Smart Cloud revealed that its Qianfan platform had officially integrated DeepSeek-R1 and DeepSeek-V3, introducing ultra-low pricing plans and offering limited-time free services through the Qianfan Model Builder.

On the same day, Alibaba Cloud announced its support for easily deploying DeepSeek-V3 and DeepSeek-R1 via the PAI Model Gallery, enhancing access for users.

On February 4, JD Cloud announced that its Yinxi AI development computing platform had begun supporting one-click deployment of DeepSeek-V3, DeepSeek-R1, alongside all distilled model variations, offering both public cloud and private instance deployment options.

On the same day, ByteDance's Volcano Engine announced it would also support various sizes of the DeepSeek open-source models.

Within the span of a month, notable companies and platforms have rallied to join the DeepSeek "friend circle."

The effects of DeepSeek's influence aren't limited to China

Advertisements

Internationally, significant cloud providers and chip manufacturers had already announced their support for the DeepSeek-R1 model earlier in JanuaryMicrosoft, for instance, announced on January 29 that DeepSeek-R1 would be available on Azure AI Foundry and GitHub, allowing developers to test the new model for cloud-based applicationsMeanwhile, on January 31, NVIDIA revealed that the DeepSeek-R1 model was accessible on the NVIDIA NIM platformBoth Amazon AWS and Microsoft Azure have since integrated the DeepSeek-R1 model.

The introduction of the open-source DeepSeek has undeniably disrupted the rhythm of development and the competitive dynamics of the AI industry both domestically and internationally.

The once notable "AI Six Tigers" (including Zero One Everything, Zhipu AI, Baichuan Intelligence, MiniMax, The Dark Side of the Moon, and Leap Star) have all begun to readjust their strategies in response to this wave of innovation.

For example, Zero One Everything and Alibaba Cloud established an "Industry Large Model Joint Laboratory," and part of the training and AI infrastructure teams joined this initiative, signaling a shift in focus from creating super large models to developing lighter, more efficient models.

On January 15, 2025, MiniMax also released and open-sourced its next-generation 01 series models, boasting a context input capacity of up to 4 million tokens—far surpassing the previous model's capacity of 128,000 tokensMeanwhile, MiniMax is examining monetization opportunities in the overseas consumer application market, with its flagship product, Talkie, emerging as a top four AI application by downloads in the United States in the first half of 2024.

Meanwhile, The Dark Side of the Moon and Zhipu AI are remaining focused on achieving technology innovations and pursuing Artificial General Intelligence (AGI), continuing to explore these commercial pathways in consumer applications.

Recent reports indicate that The Dark Side of the Moon has prioritized achieving state-of-the-art results as its crucial goal, aiming to enhance multimodal and long-text reasoning abilities in 2025.

Zhipu AI's CEO has committed to going "All in AGI," believing this path offers substantial opportunities for growth and optimization, while Baichuan Intelligence continues to focus on implementing large models within healthcare sectors.

Leap Star, meanwhile, is still in pursuit of breakthroughs in basic model research, having recently launched the Step-10 Vision multimodal understanding model, designed for end-to-end integration involving text, visual, and audio data.

As the DeepSeek model garners attention, it presents dual challenges of technological advancement and commercialization for the AI "Six Tigers." Moving forward, new participants in the AI landscape will likely pursue more segmented business paths.

Advertisements

Advertisements