Conceptual diagram of AI Photo: VCG

As MiniMax launches a new artificial intelligence (AI) agent designed to handle a wider range of visual creation tasks, a broader shift is taking shape across China's AI industry, with leading companies opening a new front in the race to capture user entry points and drive user growth.

Industry experts said that as leading AI models become increasingly capable and the focus shifts from generating answers to completing tasks, more established AI companies are expected to concentrate on building agents that can act on behalf of users. AI agents are emerging as a key battleground for large-language model commercialization and could become a critical gateway to the next generation of internet services.

On Thursday, MiniMax launched MiniMax Design, a multimodal creative agent workspace aimed at further unlocking the productivity of its general-purpose multimodal generation model H3, according to the company's official WeChat account.

The agent is built around H3, enabling it to handle the entire workflow from asset processing and content generation to editing and final delivery.

MiniMax Design embeds its understanding of H3's capability boundaries, input requirements and usage methods into its agent and skills. For a given task, the system determines what text, images, video and audio are needed, which materials should serve as references, and what should be retained, enhanced or modified, before translating the creative request into a combination of assets, instructions and capabilities optimized for H3, according to the company's post.

On August 13, DeepSeek open-sourced DeepSeek Harness, an AI agent development framework, on AIHub. Built around an "everything is a plug-in" architecture, the framework implements core agent capabilities including models, tools, skills, sessions, sandboxes, storage, loops, and scheduling as plug-ins, allowing developers to replace or extend any layer without modifying the framework's source code.

The shift by Chinese AI companies from focusing primarily on foundation models toward agents and real-world applications is a natural step toward commercialization, Xiang Ligang, director-general of the Zhongguancun Modern Information Consumer Application Industry Technology Alliance, told the Global Times on Thursday. The key value of agents lies in turning model capabilities into practical execution capabilities that can be integrated into real world use cases, he added.

On July 27, Chinese AI agent Intelligence-Indeed Agent topped the global OSWorld leaderboard with a task success rate of 90.2 percent, also taking first place on the agent-specific ranking. Its score surpassed the highest publicly reported results from Silicon Valley players including Meta, Anthropic and OpenAI.

According to industry data released in early August, China's AI agent market is projected to expand from 2.6 trillion yuan ($386.9 billion) in 2025 to 20.1 trillion yuan by 2030, with a compound annual growth rate of 50.4 percent. The business segment is expected to be the main growth driver, contributing about 90 percent of revenue over the long term.

China's advances in AI agents show that competing with Silicon Valley's computing and capital advantages requires strength in engineering and deployment in the real world, a Chinese expert noted.

Ultimately, the success of agent commercialization will depend first on how well companies understand and address genuine user needs. As AI becomes more deeply integrated with consumer devices, capabilities in hardware development, manufacturing and system integration could also become increasingly important, Xiang said.