Liang Wenfeng Bets on AI Agents, DeepSeek Plans to Release Agent Model by Year-End, ‘Agent Wars’ Set to Begin

Liang Wenfeng Bets on AI Agents, DeepSeek Plans to Release Agent Model by Year-End, 'Agent Wars' Set to BeginImage Source: DeepSeek

According to a report by Bloomberg, DeepSeek is developing a new large model with enhanced autonomy, aiming to compete directly with American giants like OpenAI in the next phase of the AI Agent competition.

Unlike traditional large models that rely on repeated user inputs and commands, this new model is designed to represent users in completing multi-step tasks with “minimal prompts” and to continuously learn and optimize during execution. Insiders indicate that this strategic shift marks DeepSeek’s preparation to advance its product to the form of a “true agent,” representing the next stage of artificial intelligence development.

DeepSeek founder Liang Wenfeng plans to officially launch this new software in the fourth quarter of this year. This timing is crucial: the entire industry is still waiting for the successor to R1, which made a splash earlier this year by performing comparably or even surpassing OpenAI in benchmark tests, despite its “million-dollar cost.”

The debut of R1 positioned DeepSeek as one of the most disruptive AI newcomers in China. However, in the following months, DeepSeek has only released minor iterations, while competitors in China and the U.S. have continuously introduced new models. There are two interpretations regarding the delay of R2: one is that Liang Wenfeng seeks perfection and is unwilling to rush delivery; the other is that the development process has encountered training and engineering challenges.

Industry Trend: Agents as the Next Battleground

DeepSeek’s new plan is not an isolated event but part of a broader shift among global tech giants. In recent months, OpenAI, Anthropic, and Microsoft have all launched their own Agent software, aiming to reshape work and lifestyle. Another Chinese startup, Manus AI, has also gained global attention with its concept of a “general-purpose agent.”

Compared to early chatbots that could only respond with short text, this generation of products is expected to handle more complex real-world tasks, from travel planning to code writing and debugging, truly moving towards the capability of “automated execution.”

However, most agents currently on the market still require significant human supervision, and full autonomy remains a distance away. What DeepSeek aims to do is to minimize the need for such “adult supervision,” allowing AI to truly become a reliable agent of action.

Chinese Landscape: Giants Move Fast, DeepSeek Moves Steadily

Compared to local giants like Alibaba and Tencent, DeepSeek’s development pace is notably restrained. Alibaba’s Qwen series models are rapidly accumulating users, and Tencent is frequently releasing new versions. In contrast, DeepSeek has maintained a low profile and restraint, but this sense of pacing highlights Liang Wenfeng’s strategic choice: to avoid unnecessary frequent updates before the new form of agents matures, concentrating resources on products that could genuinely change the industry landscape.

Whether DeepSeek can ignite global discussions again like R1 remains uncertain. However, it is certain that this company is betting its future on the “AI Agent” track, which is also the battleground for the next decisive contest in the entire artificial intelligence industry.

Source:

[1] https://www.bloomberg.com/news/articles/2025-09-04/deepseek-targets-ai-agent-release-by-end-of-year-to-rival-openai?srnd=homepage-asia&embedded-checkout=true

Source: Z Finance Statement: The purpose of this public account (ID: czfida) is to disseminate more information, and the copyright belongs to the original author. It is not for commercial use. If there is any infringement, please contact us.

Liang Wenfeng Bets on AI Agents, DeepSeek Plans to Release Agent Model by Year-End, 'Agent Wars' Set to Begin

Liang Wenfeng Bets on AI Agents, DeepSeek Plans to Release Agent Model by Year-End, 'Agent Wars' Set to Begin

Zhongguancun Fintech Industry Development Alliance

Zhongguancun Fintech Industry Development Alliance

Liang Wenfeng Bets on AI Agents, DeepSeek Plans to Release Agent Model by Year-End, 'Agent Wars' Set to Begin

The Zhongguancun Fintech Industry Development Alliance (referred to as “Zhongjin Alliance”) is a community organization approved by the Ministry of Civil Affairs, guided by the Beijing Municipal Science and Technology Commission, the Zhongguancun Management Committee, and the Haidian District People’s Government. It was jointly initiated by leading fintech companies and professional service institutions, including the Zhongguancun Internet Finance Research Institute, China Science Software, Ant Technology, Fourth Paradigm, Ming Shi Data, Boyan Technology, UnionPay Clearing, Everbright Bank, Zhongguancun Bank, Longying Zhida, Du Xiaoman Financial, Shenzhou Information, etc. The alliance currently has over 300 member units, driven by innovation, centered on technology, application-oriented, and industry-focused, aiming to build a service center for the fintech industry in Zhongguancun and nationwide, establishing a cooperation platform for fintech industry-academia-research-application, and promoting the transformation and upgrading of the fintech industry.

Liang Wenfeng Bets on AI Agents, DeepSeek Plans to Release Agent Model by Year-End, 'Agent Wars' Set to Begin

Project Undertaking, Conference Cooperation, and Specialized Training

Contact Number: 18304446664

Email: [email protected]

Membership Inquiry

Liang Wenfeng Bets on AI Agents, DeepSeek Plans to Release Agent Model by Year-End, 'Agent Wars' Set to Begin

Liang Wenfeng Bets on AI Agents, DeepSeek Plans to Release Agent Model by Year-End, 'Agent Wars' Set to BeginLiang Wenfeng Bets on AI Agents, DeepSeek Plans to Release Agent Model by Year-End, 'Agent Wars' Set to Begin

Leave a Comment