Alibaba plans to require large commercial cloud providers to share revenue from hosting its flagship Qwen 3.8-Max model, shifting towards a freemium open-source strategy.
Alibaba plans to introduce a revenue-sharing requirement for large commercial enterprise users of its flagship Qwen 3.8-Max open-weight model (a 2.4-trillion-parameter MoE model). The revenue-sharing terms apply strictly to large commercial entities and cloud providers offering the model as a service at scale in their own data centres, rather than researchers, developers, or small-scale users.
Historically, Alibaba allowed open-weight Qwen models to be deployed on customers’ own infrastructure without fee under permissive Apache 2.0 licences, while only charging for models hosted via Alibaba Cloud APIs. Specific thresholds for “large commercial users” and the exact revenue-share percentage rates remain under negotiation and have not been finalised.
The policy mirrors a ‘freemium’ open-source strategy recently introduced by Chinese AI start-up Moonshot AI for its Kimi K3 model, which requires commercial providers generating over USD$20 million in annual revenue to strike a commercial deal with up to a 30 per cent revenue share.
American cloud providers like DigitalOcean, Together AI, and Fireworks AI have already established commercial agreements or partnerships around Chinese open-weight models under this emerging model. Following the news of the monetisation strategy, Alibaba stock gained 4.5 per cent in pre-market New York trading and rose 7 per cent on the Hong Kong exchange.
