Junyang Lin, the architect of Alibaba's Qwen large language model (LLM), has left the company to launch Pragmatik Labs, a startup focused on building AI agents that can operate in both digital and physical worlds. The move signals a major industry shift: as language models become commoditized, the real competitive advantage lies in creating AI systems that can actually take action, whether that means booking appointments, controlling robots, or managing complex business workflows.

Who Is Junyang Lin and Why Does His Move Matter?

Lin joined Alibaba's DAMO Academy as a senior algorithm engineer in 2019 and quickly became the technical lead for Qwen when the project officially launched in 2020. Since 2022, he focused on building generalist models, eventually creating one of China's most competitive large language models that rivals global AI giants. His work on Qwen established him as a star researcher in China's AI landscape, creating open-source models that gained significant traction in the developer community worldwide.

His departure from Alibaba to start Pragmatik Labs (also known as p7k Labs or 语用科技) in Shanghai represents a calculated bet on what comes next in artificial intelligence. Rather than competing in the crowded LLM space, Lin is pivoting toward "next-generation agents across digital and physical worlds," signaling a shift from conversational AI to more actionable systems that can interact with real environments.

What Exactly Is Pragmatik Labs Building?

The startup's focus breaks down into two complementary categories of AI agents. Digital agents handle knowledge work, operations, and industry-scale workflows that reason, use tools, and coordinate action in software environments. Physical agents represent embodied intelligence that perceives, acts, and solves long-horizon tasks in the real world, effectively bridging AI models with machines.

This dual focus addresses a fundamental limitation of today's AI systems. Language models can generate text and answer questions, but they cannot independently book appointments, control robotic arms, manage multi-step enterprise processes, or navigate physical spaces. Pragmatik Labs is positioning itself to solve that gap.

How Are Investors Betting on This Vision?

• Gaorong Ventures: Co-leading the funding round with an investment of approximately $100 million USD, signaling confidence from a top-tier Chinese venture capital firm.
• HSG (formerly Sequoia China): Co-leading with an equal $100 million USD investment, bringing the prestige of one of Asia's most influential venture firms.
• Tencent: Participating with approximately $20 million USD, demonstrating support from one of China's largest technology conglomerates.
• Shanghai Engine Fund: A state-backed future-industry fund contributing to the round, highlighting government support for the venture.

The total funding package positions Pragmatik Labs at a reported valuation of around $2 billion USD, demonstrating the immense confidence investors have in Lin's vision and track record. This level of backing from both private capital and government funds creates a formidable foundation for innovation in the AI agent space.

What Real-World Problems Could AI Agents Solve?

The practical applications of Pragmatik Labs' technology extend across multiple industries and use cases. Agents designed for booking and operations could schedule, coordinate, and complete knowledge work without requiring humans to click through every step. Physical agents could extend model intelligence into machines that move, perceive, and act in the real world, from manufacturing to logistics. Long-horizon systems could plan, use tools, and keep multi-step enterprise processes moving without human intervention.

The strategic positioning of Shanghai as Pragmatik Labs' base is particularly significant. The city serves as both a major manufacturing hub and a capital center, creating an ideal environment for developing and testing embodied AI systems that bridge software and hardware.

Why Is This Shift From Language Models to Agents Important?

Lin's transition from leading one of China's top LLM projects to founding an agent-focused startup reflects a broader industry recognition. Large language models have become increasingly commoditized, with dozens of capable options available from companies worldwide. The real differentiation and value creation now lies in systems that can autonomously perform tasks, make decisions, and interact with their surroundings.

This represents the next frontier in artificial intelligence. While conversational AI captured headlines and investment throughout the early 2020s, the companies that will dominate the next decade are those that can move beyond chat interfaces to create systems that actually do things. Pragmatik Labs, backed by Lin's proven expertise and substantial capital, is positioning itself to lead that transition.