Alibaba Group Holding has unveiled an open-source deep research AI agent that rivals OpenAI’s flagship Deep Research tool in performance, while showcasing superior efficiency, with the company claiming it outperforms larger U.S. proprietary models.
The AI agent, which has just 30 billion parameters, has been integrated into two key products, according to a blog post by Alibaba’s AI search development team, Tongyi Lab, on September 18, 2025. In the Amap maps app, users can now utilize the agent’s web retrieval capabilities to plan multi-day trips, streamlining travel itineraries with real-time data synthesis. Similarly, the AI-powered legal research tool Tongyi FaRui has been enhanced with the agent’s functions, improving its precision in retrieving case law complete with verified citations.
This release marks a significant step in Alibaba’s aggressive expansion of AI technologies. Over the past two weeks, the company introduced its first trillion-parameter base model, Qwen-3-Max-Preview, and a more compact yet potent variant, Qwen3-Next-80B-A3B, as evaluated by benchmarking firm Artificial Analysis. These developments position Alibaba as a formidable player in the global AI landscape.
Deep research agents, like OpenAI’s version launched in February and integrated into ChatGPT, are designed for intricate online information gathering. Major U.S. firms, including Google DeepMind, have followed suit with comparable tools. Alibaba’s open-source approach contrasts with these proprietary systems, potentially accelerating innovation through broader accessibility.
Alibaba, which owns the South China Morning Post, continues to invest heavily in AI to compete with international leaders. This agent not only bolsters practical applications in navigation and legal research but also underscores China’s growing influence in efficient, scalable AI solutions.
