Recently, Zhifu AI launched its autonomous intelligent body AutoGLM, similar to OpenAI's AI Agent. The AutoGLM model can accurately identify and understand user commands, without the need for users to manually demonstrate operations, not limited to simple task scenarios or API calls, can replace users in executing operations on electronic devices, can automatically complete tasks like ordering takeout, editing comments, shopping on Taobao, liking posts on Moments, summarizing article content, and other daily activities.

Zhifu AI is an artificial intelligence service launched by Peking Zhifu Huazhang Tech Co., Ltd., founded in June 2019, is a company mainly engaged in technology promotion and application services. Its core team comes from the Computer Science department of Tsinghua University, with profound technological research and development capabilities and academic backgrounds.

Institutions believe that the autonomous intelligent body AutoGLM has provided a technical paradigm for mobile AI Agents, helping users break free from tedious App operations, improving life efficiency, and is expected to truly realize the "Jarvis" becoming a reality, driving the accelerated landing of AI Agents.

Looking back at the past year, in fact, large AI models and applications in China are in a dormant state, mainly for three reasons:

1) Overseas closed-source model progress was actually lower than expected in the early stages. This year focused on open-source models, and open-source models gradually caught up with the capabilities of closed-source models this year. However, according to the latest media reports, OpenAI's latest model is expected to be released within the year.

2) Domestic models are still in a catching-up state. In fact, this year has not significantly exceeded the expectations for domestic models at the beginning of the year.

At the global level, there has not yet been a new all-encompassing blockbuster similar to GPT in terms of application.

Based on these three reasons, in fact, this year, the large-scale models and AI applications in the domestic A-share market may be in a dormant state, but institutions believe that AI Agent may be a direction worth paying attention to at this point.

Open AI has actually launched its AI application market, but the past applications have been relatively flat.

As Apple, Samsung, and Huawei gradually implement AI Agents, especially the cooperation between leading terminal smartphone manufacturers and leading model manufacturers, the penetration of AI applications in mobile terminals will accelerate. AI Agent is expected to usher in a faster wave of promotion and application compared to the past year, which may also impact the entire terminal manufacturers, including model manufacturers' competitive landscape.

China International Capital Corporation published a research report stating that since 2024, AI video tools have undergone multiple rounds of technological function upgrades and gradually started the commercialization process. DiTs architecture is currently mainstream; these tools are being implemented in practical applications for content production.

China International Capital Corporation believes that AI video tools are expected to first achieve cost reduction and efficiency improvement in the professional content creation field, revolutionizing the video content production process gradually. Acceptance at the consumer end is also increasing, with a core focus on breaking the circle of AI video content, but attention should still be paid to issues such as the match between existing UGC content preferences and AI video features, video generation technology iterations, and consumer application norms (such as copyrights).

AI video related Hong Kong stocks:

Kuaishou-W (01024): AI video applications integrated into short video production processes, driving the transformation of AI + short videos.

Bilibili-W(09626): AI content empowers gaming and advertising, already has self-developed vertical model, technology advantages waiting for further exploration.