Chinese AI giants pivot toward proprietary models to drive revenue, performance

Major Chinese AI players, including Alibaba Cloud and Zhipu AI, are strategically pivoting toward keeping their most advanced models proprietary to drive revenue and performance. This week, Alibaba released three new models—Qwen3.6-Plus, Qwen3.5-Omni, and another unspecified model—that are accessible only through its official cloud platform or chatbot website. This represents a significant shift, as the previous generation, Qwen3-Omni released in September 2025, was open-sourced. An Alibaba Cloud spokesperson cited lower developer download rates on Hugging Face for the Omni series as a reason for keeping the latest version closed.

The move signals a broader industry trend where the computational and hosting demands of state-of-the-art models make local deployment impractical, pushing companies toward cloud-based, revenue-generating APIs. While companies maintain they are not abandoning open-source entirely, the most powerful and capable models are increasingly being reserved for commercial channels. This allows firms to capture the full value of user traffic and computational usage, turning research investments into direct revenue streams. The strategy highlights the growing monetization pressure in the high-stakes AI race, even as open-source communities continue to play a vital role in ecosystem development.