Alibaba releases a fully open source AI agent that rivals top US models, offering high performance with fewer parameters and greater accessibility for developers.
Alibaba Group Holding has unveiled a leading open source deep research AI agent, marking a significant step in accessible and transparent AI development. Unlike proprietary US tools, this agent allows developers and researchers to access, modify, and deploy a high-performance AI model without restrictions, lowering barriers to adoption and fostering innovation.
Despite operating with only 30 billion parameters—far fewer than typical US models—the agent reportedly matches the performance of OpenAI’s flagship deep research tool, demonstrating “incredible efficiency,” according to Alibaba. This streamlined design enables complex, multi-step web retrieval tasks while maintaining high computational efficiency.
The agent is already being integrated into Alibaba’s ecosystem. In the Amap app, users can leverage its web retrieval capabilities to plan multi-day trips efficiently. Meanwhile, the Tongyi FaRui AI legal research tool benefits from enhanced case law retrieval with verified citations, demonstrating the agent’s versatility across applications.
This release follows Alibaba’s broader AI expansion, including the launch of Qwen-3-Max-Preview, the company’s first trillion-parameter base model, and Qwen3-Next-80B-A3B, a smaller yet more powerful model benchmarked by Artificial Analysis. Globally, deep research agents have become a focus, with OpenAI introducing one in February via ChatGPT and US companies like Google DeepMind developing similar tools.
Alibaba’s open source approach not only challenges the dominance of US proprietary AI models but also sets a precedent for efficient, transparent, and widely accessible AI research, proving that fewer parameters can still deliver cutting-edge performance.
