The realm of AI large models is hardly a stranger to anyone. Since GPT ushered in the era of large models, these sophisticated systems have profoundly transformed our lives. However, starting this year, with the emergence of DeepSeek, the once-stabilizing large model market has reignited with fervor. DeepSeek even sparked a countertrend, followed by the recent debut of Manus, which once again stirred up market turbulence. What’s driving this renewed AI melee?

I. The Evolution of the Large Model Market: From DeepSeek to Manus

According to a China News Weekly report, the rise of DeepSeek has birthed the first wave of beneficiaries in the traffic feast.

Tencent Yuanbao, an AI assistant under Tencent, surged past DeepSeek to claim the top spot in the free App download ranking on the Chinese Apple App Store. Following its integration of the DeepSeek-R1 model on February 18, Tencent Yuanbao ascended from obscurity to the pinnacle in less than 20 days.

On January 20, DeepSeek unveiled the large model DeepSeek-R1. Besides offering cost-effectiveness and robust reasoning abilities, DeepSeek’s free open-source model overturned the established rules of the large model market. Subsequently, major domestic companies accelerated their adoption of large models, with players like Baidu, Tencent, ByteDance, Alibaba, and JD.com integrating DeepSeek into their respective systems.

Behind DeepSeek’s pursuit of AI equality lies a fresh wave of traffic opportunities. Amidst the bottleneck in Internet traffic, large companies are ramping up their efforts to gain an edge in the competition for user attention and engagement.

Shortly thereafter, on the late night of March 5, Beijing time, another "lightning strike" reverberated through the AI landscape. An AI product named Manus made its debut, introducing itself as the "first general AI Agent" capable of independent thinking, planning, executing complex tasks, and delivering complete results.

Its promise to genuinely assist people in task execution has attracted widespread attention, with many eager to try it out. However, as Manus was still in internal testing, most failed to secure an invitation code despite staying up all night. Those who awoke to find Manus trending were astounded by invitation codes being hyped up to several hundred or even tens of thousands of yuan on second-hand platforms, leading to anxious waits for the opportunity to try it out.

Moreover, as a "phenomenal work" from a Chinese team, it’s no surprise that netizens dubbed this night "another sleepless night in the tech circle after DeepSeek".

Public information reveals that Xiao Hong, the founder of Monica.im, the team behind Manus, graduated from Huazhong University of Science and Technology with a degree in software engineering and is a serial entrepreneur. Xiao Hong has successively launched products such as WeBan, a public account editor for B-end users; WeBan Helper, a private domain traffic management platform; and Monica, a browser plugin integrating multiple large models to provide chat, translation, and other functions for C-end users.

II. The Catalysts Behind the Renewed AI Melee

As we step into 2025, the AI landscape is evolving rapidly, and a fresh melee has quietly commenced. How should we perceive this AI melee?

Firstly, DeepSeek’s rise truly broke the stalemate in the large model market. Since its widespread adoption this year, DeepSeek’s open-source feature has been a pivotal factor in its rapid market expansion. Previously, large models were often closed off, meaning only a few enterprises with robust technology and resources could build and optimize them. This exclusivity fostered a technological monopoly, somewhat stifling AI development.

DeepSeek’s open-source approach shattered this monopoly. For developers, open-source is akin to unlocking a treasure trove. They can freely access, modify, optimize, and innovate upon DeepSeek’s code according to their needs. This enables more small and medium-sized teams and researchers to participate in AI development, no longer hindered by resource and technological thresholds. From a market perspective, this is a stride towards AI equality, as AI technology once monopolized by a few giants begins to disseminate to a broader audience, invigorating the entire market.

Secondly, the anticipation surrounding AI innovations like DeepSeek has fully ignited the capital market. Both DeepSeek and Manus embody the market’s keen anticipation for impactful AI products. In today’s technological era, AI is viewed as a field with immense potential capable of bringing unprecedented changes. Investors, enterprises, and consumers alike are eagerly awaiting the next AI product with a revolutionary impact akin to the Internet revolution.

When DeepSeek demonstrated its unique advantages through open-source, the market recognized its potential to become a blockbuster product. Similarly, if Manus announces a significant breakthrough, it will garner substantial market attention. This interest stems not only from curiosity about the technology but also from the substantial commercial benefits it holds.

The market is highly responsive to news of seemingly groundbreaking products. Whenever such products emerge, there’s significant activity in the capital market. This is because the AI market holds immense potential, and all parties aim to seize the initiative in this burgeoning field. Manus’s timely emergence, coupled with its positioning as an "actively working" AI, aligns with the capital market’s appetite for AI narratives, thereby igniting a confluence of technology and traffic. Consequently, "DeepSeek or Manus concept stocks" have emerged at this historic juncture.

Thirdly, the world anticipates the genuine arrival of AI’s "iPhone moment". The current market yearns for AI’s equivalent of the iPhone moment. Looking back, the advent of the iPhone revolutionized the mobile phone industry and redefined the mobile Internet era. In AI, while current large model products have made strides, most are still confined to simple dialogue scenarios, far from seamlessly replacing human work.

Recalling the iPhone’s impact, we believe that for an AI super application to achieve iPhone-like success, it must embody the following traits:

First, it must exhibit a high degree of intelligence and autonomy, capable of thinking, learning, and decision-making akin to humans, adjusting its behavior and strategy according to diverse scenarios and needs. For instance, in industrial production, a super AI could autonomously monitor production parameters, promptly detect issues, and make adjustments to achieve automation and intelligence.

Second, it must have a vast array of application scenarios, spanning various fields and industries, bringing about comprehensive changes to people’s lives and work. For example, in education, a super AI could offer personalized learning plans based on students’ conditions and characteristics; in transportation, it could enable autonomous driving and intelligent traffic management, enhancing efficiency and safety.

Third, it must offer a stellar user experience, featuring simple operation, a user-friendly interface, and ease of use for ordinary users. Just like the iPhone, its success lies not just in advanced technology but also in its intuitive interface and human-centric design, making it accessible and enjoyable for a vast user base.

However, we’re still a long way from achieving this vision. Yet, market expectations propel major enterprises and developers to continuously invest resources in AI technology advancement, which is also a significant driver behind the AI melee.

Fourthly, the AI melee harnesses the law of survival of the fittest to foster more innovations. The AI melee is not detrimental; rather, it’s beneficial. Human technological history tells us that ample competition is crucial for achieving substantial development. Previous technological revolutions, marked by fierce competition, have driven rounds of technological breakthroughs. For instance, during the Industrial Revolution, numerous enterprises competed in steam engine and textile technologies, continually enhancing them and boosting production efficiency, thereby propelling rapid industry growth. Similarly, in the early stages of the Internet’s development, fierce competition among Internet enterprises in search engines, social networks, e-commerce, and other domains gave birth to giants like Google, Facebook, and Amazon, transforming how people access information, socialize, and shop.

In AI, competition spurs enterprises to invest in research and development, enhance model performance, reduce costs, and expand application scenarios. DeepSeek broke the hegemony of computational power through technological innovation, while Manus conducted engineering optimizations at the application level to meet the market’s demand for practical AI. Major technology companies have also intensified their AI layout, launching their models and applications. This competition fosters rapid technology iteration and innovation, accelerates AI technology’s maturity and popularization, and ultimately propels the AI industry to new heights. Only the genuine arrival of AI’s "iPhone moment" can transform the current market landscape.

Therefore, we should embrace this melee phenomenon and anticipate the emergence and development of more exceptional AI products in the future.