Lin Junyang, the former technical lead of Alibaba Group Holding's Qwen model family, has founded Pragmatik Labs in Shanghai with a $2 billion valuation at angel stage1. The company has no announced product, user base, or revenue.
Gaorong and HongShan co-led the round with $100 million each, Tencent invested $20 million, and the Shanghai Future Industry Fund also participated. Lin personally controls most of the equity through related entities, with external investors holding 12 percent of the domestic operating company, which operates under the Chinese name Yuyong Technology.
Pragmatik Labs is publicly framed as a next-generation agent company spanning digital and physical worlds. Its website describes work on reasoning, tool use, environment interaction, and long-horizon task execution, with two business lines: digital agents for knowledge work and physical agents for real-world environments.
Lin's career arc underpins the valuation. He joined Alibaba DAMO Academy in 2019 after completing his master's degree, working on the M6 and OFA pre-training models before becoming Qwen's technical lead in late 2022. He was one of the youngest P10-level engineers at Alibaba Group Holding. Qwen launched its first open-source model in August 2023 and expanded into a family covering language, code, vision, and multimodal capabilities. By January 2026, more than 200,000 Qwen-derived models existed globally.
While still at Qwen, Lin set up a robotics and embodied intelligence team in October 2025. He resigned on March 4 and published a long-form essay on March 26 titled "From Reasoning Thinking to Agentic Thinking".
The $2 billion angel-stage valuation — reached without a product — reflects a pattern at the top of the AI talent market, though it remains notable inside China. The report notes that Thinking Machines Lab, founded by Mira Murati after leaving OpenAI, raised $2 billion at a $12 billion valuation in early stages as a comparable data point.
Tencent's $20 million check in Pragmatik Labs arrives as the company is sharply accelerating AI spending; Tencent nearly tripled its capital expenditure in the second quarter of 2026 to support its AI buildout across models, applications, and infrastructure ctx.
ANALYSIS The pivot from foundation-model development to agent-focused work is the strategic signal. Lin led one of the most widely adopted open-source model families in the world, yet chose to build a new company around agents — digital and physical — rather than continue competing on base-model scale. The framing of Pragmatik Labs around reasoning, tool use, and long-horizon task execution positions the company downstream of foundation models, building on top of the capabilities Lin helped create at Qwen.
The investor syndicate — pairing top-tier Chinese venture firms Gaorong and HongShan with a strategic check from Tencent and municipal participation via the Shanghai Future Industry Fund — concentrates financial, distribution, and policy support in a single cap table. Lin's retention of majority equity through related entities is an atypical structure for a Chinese AI startup at this valuation level, suggesting strong founder leverage in the negotiation.