Alibaba Unveils Open-Source AI Agent Challenging OpenAI’s Deep Research with Unmatched Efficiency

In a landmark moment for artificial intelligence development, Alibaba Group Holding has introduced an open-source deep research AI agent that rivals the capabilities of OpenAI’s flagship Deep Research tool, marking a significant stride in accessible, transparent, and efficient AI technology. Announced in September 2025, Alibaba’s agent demonstrates remarkable performance with only 30 billion parameters, a fraction of those typically used by U.S. proprietary models, signaling a breakthrough in achieving cutting-edge AI with vastly reduced computational demands.

Background: The AI Research Landscape and Deep Research Agents

Deep research agents represent a new class of AI systems designed to conduct complex, multi-step web retrieval and information synthesis tasks autonomously. These agents excel at pulling together detailed data from multiple online sources, evaluating relevancy, verifying facts, and supporting high-value decision-making across industries.

OpenAI set the benchmark earlier this year by integrating its Deep Research agent within ChatGPT, enabling users to perform rigorous, context-aware web research beyond conversational capabilities. Similarly, Google DeepMind and other leading technology firms have pursued comparable innovations focusing on transparency, accuracy, and operational efficiency.

Alibaba’s Strategic Open-Source Approach

Alibaba’s newly released agent, Tongyi DeepResearch, embodies the company’s commitment to democratizing AI access through an open-source model. Unlike proprietary tools restricted to closed ecosystems, Tongyi DeepResearch allows developers worldwide to access, modify, and deploy a powerful AI research assistant freely, fostering community-driven innovation and accelerating AI adoption.

This openness contrasts notably with U.S.-based AI research tools that often involve high costs, black-box operation, and limited customization. Alibaba’s move aims to lower the barriers to entry for advanced AI technologies, enable broad collaboration, and cultivate an ecosystem where AI tools evolve more rapidly and responsively to users’ needs.

Sponsored Banner

Technical Excellence and Efficiency

Technical Excellence and Efficiency

Powered by 30.5 billion total parameters, with selective activation of only 3.3 billion parameters per inference token, Tongyi DeepResearch achieves exceptional computational efficiency without compromising accuracy. Parameters—variables defining AI model intelligence—are usually correlated with performance; however, Alibaba demonstrates that strategic design can yield superior outcomes using fewer resources.

The AI agent excels in rigorous benchmark tests, including Humanity’s Last Exam, which challenges AI systems with complex academic problems requiring multi-step reasoning. Alibaba’s agent scored 32.9%, surpassing OpenAI’s Deep Research score of 26.6% recorded earlier in the year. Such results highlight the model’s sophisticated reasoning capabilities despite a smaller size.

Integration and Versatility in Real-World Applications

Alibaba has incorporated Tongyi DeepResearch into key products within its ecosystem. In the Amap application, users benefit from advanced web retrieval capabilities to efficiently plan multi-day trips, integrating geographically and temporally complex data. Meanwhile, the Tongyi FaRui AI legal research platform employs the agent to enhance case law retrieval with verified citations, demonstrating practical utility in sectors like law.

These integrations showcase the model’s flexible adaptability across use cases, from consumer apps to professional services, enhancing user empowerment through precise, real-time data retrieval and reasoning.

Open-Source Community Reception and Future Development

Following its open-source release on platforms such as Hugging Face, Tongyi DeepResearch has seen rapid adoption by the global developer community. Enthusiasts and AI researchers actively experiment with and refine the model, unlocking novel functionalities and contributing to continual improvement.

Adina Yakefu, a machine learning community manager at Hugging Face, praised the model’s “impressive” benchmark performance and anticipated its popularity would spur collaborative innovation, accelerating progress in AI agent capabilities worldwide.

Alibaba continues to invest in advancing Tongyi DeepResearch through ongoing pre-training on diverse, high-quality synthetic datasets and reinforcement learning approaches designed to stabilize performance in dynamic environments. These efforts underscore Alibaba’s long-term vision for sustained AI leadership driven by transparency, collaboration, and efficiency.

Sponsored Banner

Broader Industry and Geopolitical Implications

Alibaba’s successful open-source launch signals a notable shift in the global AI landscape, challenging the dominance of U.S.-proprietary models exemplified by OpenAI. By proving that fewer parameters can deliver state-of-the-art performance with increased efficiency, the company sets a precedent for more sustainable, accessible AI innovation.

This move aligns with China’s broader ambitions to establish technological self-reliance and global leadership in AI, incentivizing other enterprises to pursue open AI research collaborations and contribute to a balanced, diverse international AI ecosystem.

The competitive yet cooperative dynamic between Alibaba and Western counterparts promises a healthier, more inclusive AI future where breakthroughs arise not from closed competition but community-driven excellence.

Challenges and the Road Ahead

Despite impressive achievements, challenges remain. Ensuring robustness against adversarial attacks, safeguarding data privacy, navigating complex ethical considerations, and meeting diverse regulatory requirements loom large as AI agents proliferate.

Alibaba will need to maintain rigorous quality control while supporting an open innovation model, balancing speed of development with safety and accountability. Continued engagement with international standards bodies and AI ethics organizations will be pivotal.

A Defining Moment for AI Accessibility and Innovation

Alibaba’s introduction of the open-source Tongyi DeepResearch AI agent heralds a transformative chapter in artificial intelligence. By matching and exceeding the performance of leading proprietary models while operating with far fewer parameters, the agent proves that efficiency and openness can coexist with excellence.

This development expands the possibilities for AI integration across global industries, democratizing access to sophisticated research tools and empowering developers, businesses, and individuals alike.

As the AI community embraces this new era of collaborative innovation, Alibaba’s move highlights the potential for technology to transcend borders, accelerate progress, and shape a more equitable AI-driven future for all.