Alibaba Unveils an Open-Source AI Agent — a 30-Billion-Parameter Challenger to OpenAI’s Deep Research
Alibaba has released an open-source deep research AI agent designed to rival OpenAI’s flagship Deep Research. In demonstrations, Amap users can leverage the agent’s web-retrieval capabilities to plan multi-day trips. Separately, Tongyi FaRui has been updated with the agent’s research functions, improving its ability to retrieve case law with verified citations. Alibaba says the deep research agent delivers “incredible effi ciency” compared with US proprietary tools, running on just 30 billion parameters—far fewer than the counts estimated for the models behind US deep-research agents.
In This Article:
What Deep Research Agents Do: Multi-Step Web Retrieval
Deep research agents are AI tools built to perform multi-step web retrieval and data gathering. OpenAI’s Deep Research was the first of this kind, launched and integrated into ChatGPT in February. Since then, other major US tech companies, including Google DeepMind, have introduced similar tools.
Real-World Use: From Amap to Law: Tongyi FaRui
Among everyday users, the technology can assist with practical tasks. In Amap, the deep research agent’s web retrieval helps plan multi-day trips. Tongyi FaRui has also been updated with the agent’s research functions, boosting its ability to retrieve case law with verified citations.
A Global AI Race: OpenAI, Google, and the Case for Efficiency
This development sits within a broader AI race. OpenAI launched Deep Research in February and integrated it into its offerings, with Google DeepMind and others following with similar tools. Alibaba’s claim of “incredible efficiency” at just 30 billion parameters challenges the assumption that only huge models win, highlighting the potential of a lean, open-source approach.
Why It Matters and What Comes Next
Why this matters: open-source AI could democratize access to powerful retrieval tools, accelerate development, and increase transparency. Yet questions remain about reliability, citation verification, and governance. The trajectory suggests a shift toward more accessible AI tools that still deliver strong performance.
