Junyang Lin, the former architect of Alibaba's Qwen large language model, has founded Pragmatik Labs and secured $2 billion in funding at a post-money valuation, according to the company announcement on August 12. The round was co-led by HongShan and Gaorong Ventures, with participation from Tencent Holdings and Shanghai Future Industry Fund.
Lin's departure from Alibaba and immediate backing from major Chinese venture investors and the Tencent unit occurred in a competitive market for talent in domestic AI. Qwen, which Alibaba has deployed across its cloud business and consumer platforms, ranks among China's most widely used models alongside ByteDance's Doubao and other rivals.
Lin directed the development of Qwen as Alibaba's chief AI scientist before leaving to start Pragmatik Labs. Qwen competes directly with models offered by ByteDance, Baidu and others in a market where the largest Chinese tech firms have invested billions in training and inference infrastructure. The new venture focuses on AI agents, a category that has attracted capital from OpenAI, Anthropic and others globally.
HongShan and Gaorong Ventures, which co-led the round, have backed other AI infrastructure and model makers in China. Tencent's investment is part of the company's strategy to fund external AI research alongside its internal teams. Shanghai Future Industry Fund, a city-backed vehicle, has supported other deep-tech startups in the region.
The $2 billion valuation at founding is typical for AI researchers with established track records launching startups in China's market. Lin's role in building one of China's most deployed models provides proof of execution that investors typically require before backing new ventures at this scale. Investors committed this amount based on Lin's demonstrated ability to build a competitive AI systems company.
AI agent startups globally have raised capital at multibillion-dollar valuations since late 2024, though most remain pre-revenue. Pragmatik Labs' backing places it alongside other well-funded agent-focused teams but does not yet confirm commercial traction or model performance against existing systems. The metric that will matter is whether Pragmatik Labs ships agents that customers adopt at meaningful scale within 18 months of launch.
