Skip to content
Impulse TeamsImpulse Teams

News

Open AI Weights Are Testing Revenue Sharing

August 17, 2026

Grainy gold light dividing across a dark abstract field with a centered OPEN VALUE label

Open weights have usually come with one of two economic stories. A permissive license lets others build freely, or a vendor keeps the strongest model closed and charges for API usage. Alibaba may be testing a third route.

Reuters reporting summarized by AI News says large commercial users of a future Qwen model may need a revenue-sharing agreement. The rate and final terms were not published. Current Qwen3 models remain under Apache 2.0, so this is a reported plan, not a retroactive change to every Qwen release.

Open distribution does not remove the revenue question

Training and serving frontier models costs money. Permissive releases can grow an ecosystem and increase demand for cloud, chips, tooling, and services, but they do not always return value directly to the model builder.

Revenue sharing attempts to connect commercial success downstream with investment upstream. It can keep weights available while asking the largest businesses to pay when they build meaningful revenue on top.

The definition of open becomes decisive

The risk is ambiguity. A model can publish weights while placing material restrictions on how successful products use them. If thresholds, attribution rules, audit rights, or revenue definitions are unclear, teams cannot price the dependency or judge whether the license will survive growth.

Traditional closed APIs are simpler in one respect: the cost is tied to usage and the provider controls the service. Permissive open source is simpler in another: the license does not change when the adopter succeeds. Revenue sharing adds flexibility, but also commercial negotiation and reporting.

A useful third model needs predictable rules

This structure could fund open-weight development without forcing every serious model behind an API. It could also give smaller builders broad access while asking only large commercial deployments to share value.

Before adoption, teams need to know what triggers payment, how revenue is measured, which products are covered, and what happens after a merger or rapid growth. Open weights can support a broad ecosystem, but businesses will build on them with confidence only when the economics stay predictable.

Sources

Ready to build your own update?

Tell us your current blockers and desired outcomes. We will propose a practical first execution scope.