Alibaba tests revenue-sharing business model for Qwen open-source AI

Alibaba's experiment with revenue sharing for Qwen tackles the central tension in open-weight AI: how to build a business around models you give away. Per Artificial Intelligence News, Alibaba is testing a model where value — and revenue — is shared, an attempt to capture upside from the wide adoption Qwen already enjoys without simply locking the weights behind a paid API.

The context is China's intensifying cost race. Alibaba and DeepSeek have driven Chinese model pricing sharply down, and this week DeepSeek conceded that ultra-low pricing was unsustainable by warning of a price hike. Alibaba's revenue-sharing test is a different answer to the same problem: rather than raise prices, find alternative monetization that keeps the models open and widely deployed.

Strategically, Qwen has become one of the most-used open-weight families globally, and Alibaba's own account teased Qwen3.8-Max and referenced an upcoming Qwen 3.8 drop, with NVIDIA's vLLM team noting optimizations for the line. Wide adoption is the asset; the question is converting it to durable revenue. A revenue-sharing scheme — potentially with hosts, integrators, or downstream commercial users — could align incentives across the ecosystem.

The skeptical read is that open-weight monetization has repeatedly proven hard, and 'revenue sharing' is vague until terms are public. If Alibaba leans too hard on monetization, it risks pushing developers toward truly permissive alternatives like Mistral's Apache-2.0 releases; too little, and Qwen stays a cost center. Watch for concrete terms and whether other open-weight vendors adopt similar models as the pricing war cools.