A professional works at home with a computer displaying a futuristic AI agent interface, reflecting DeepSeek’s push to develop advanced autonomous AI systems. Image Source: ChatGPT-5
China’s DeepSeek to Launch AI Agent by 2025 to Rival OpenAI
Key Takeaways: DeepSeek’s Push Into AI Agents and Global Competition
DeepSeek, a China-based startup, is building an AI agent designed to execute multi-step tasks with minimal user input.
The new system will also be able to learn and improve from its prior actions.
Founder Liang Wenfeng is pushing to release the agent by Q4 2025, even as the industry awaits DeepSeek’s R2 model.
DeepSeek’s earlier R1 model, launched in January, rivaled OpenAI on benchmarks while costing only millions to build.
U.S. players like OpenAI, Anthropic, and Microsoft and Chinese rivals such as Alibaba and Tencent are all racing to develop agentic AI.
DeepSeek’s Next Ambition: Building an AI Agent
DeepSeek is developing an artificial intelligence model with advanced agent features intended to rival U.S. companies such as OpenAI. According to people familiar with the matter, the China-based startup is designing the system to perform multi-step actions on behalf of users with minimal direction.
The agent will also be built to learn from its own prior actions and improve over time, signaling a step toward greater autonomy in AI systems. The planned release is expected in the final quarter of 2025, according to the people, who requested anonymity because the information is private.
From R1 Success to Delayed R2 Rollout
The announcement comes as the AI industry is still digesting the impact of DeepSeek’s earlier R1 model, which launched in January. That platform, designed to mimic human reasoning processes, cost only several million dollars to build but managed to match or surpass OpenAI’s products in benchmark tests.
Since then, DeepSeek has released only minor updates, while rivals in both China and the U.S. have rapidly pushed out new models. Local media have attributed the delay of a follow-up R2 model to founder Liang Wenfeng’s determination to “get it right,” while others have speculated about potential training or development glitches.
Liang continues to run High-Flyer Asset Management, his investment firm, alongside DeepSeek. The company did not respond to requests for comment.
AI Agents: The Next Frontier in Artificial Intelligence
The push toward AI agents is part of a broader industry trend. Companies such as OpenAI, Anthropic, and Microsoft have recently launched their own agent software, intended to help users streamline both personal and professional tasks.
Rather than providing only short text responses like earlier chatbots, these systems are being designed to manage complex workflows, from planning travel itineraries to writing and debugging computer code. A recent Goldman Sachs research report identified AI agents as the next landmark advance in generative AI, with the potential to boost business productivity significantly.
Competition Inside and Outside China
While many Chinese firms are moving aggressively into AI, DeepSeek has kept a relatively muted pace. Competitors such as Alibaba Group Holding Ltd. and Tencent Holdings Ltd. have rolled out a steady stream of models and upgrades. Alibaba’s Qwen series in particular has gained strong popularity among developers and businesses.
By contrast, DeepSeek’s agent-focused project underscores a strategy of focusing on breakthrough capabilities rather than frequent updates. The question remains whether the company can replicate the disruptive success of the R1 model as it prepares to unveil its next-generation AI agent in 2025.
Q&A: DeepSeek’s Planned AI Agent
Q: What is DeepSeek developing?
A: DeepSeek is building an AI agent that can perform multi-step tasks with minimal user input and learn from its past actions.
Q: When will the new AI agent be released?
A: The company is aiming for a Q4 2025 launch.
Q: How did DeepSeek’s R1 model impact the industry?
A: The R1, released in January, rivaled OpenAI’s products on benchmarks despite costing only millions to build.
Q: Why has the R2 model been delayed?
A: Reports suggest Liang Wenfeng wanted to ensure quality, while others point to possible development or training issues.
Q: Who are DeepSeek’s main competitors in AI agents?
A: In the U.S.: OpenAI, Anthropic, Microsoft. In China: Alibaba’s Qwen models and Tencent’s AI systems.
What This Means: China’s Entry Into the AI Agent Race
DeepSeek’s decision to prioritize an AI agent for late 2025 underscores how quickly the industry is shifting from conversational chatbots to systems that can initiate, execute, and learn from complex tasks. If successful, this could move artificial intelligence closer to acting as a true digital assistant, capable of driving business productivity and reshaping knowledge work.
The strategy also illustrates China’s ambition to challenge U.S. leadership in advanced AI. By aiming to compete directly with OpenAI, Anthropic, and Microsoft, DeepSeek is positioning itself at the center of a geopolitical technology race where breakthroughs carry both economic and strategic weight.
Yet the company faces a dual challenge: maintaining the disruptive reputation it earned with the R1 model, while proving it can deliver another leap forward after delays around the R2. With rivals like Alibaba and Tencent flooding the market with frequent releases, DeepSeek’s more measured pace raises the stakes for its agent launch.
The outcome will help determine whether DeepSeek becomes a lasting global player in agentic AI or risks being overshadowed by competitors moving faster to define the future of autonomous systems.
Editor’s Note: This article was created by Alicia Shapiro, CMO of AiNews.com, with writing, image, and idea-generation support from ChatGPT, an AI assistant. However, the final perspective and editorial choices are solely Alicia Shapiroo’s. Special thanks to ChatGPT for assistance with research and editorial support in crafting this article.