Of all the strong open-weights releases, Qwen has one of the cleanest licensing stories. Alibaba has released many Qwen checkpoints under Apache 2.0. The permissive, well-understood one. For most SaaS products, that is about as clean as open-weights licensing gets.
The catch, as always: check your specific version. The Qwen family is big, and licensing has varied across releases and sizes. Do not assume one release's license applies to the next. Verify the license file attached to your exact checkpoint. Under Apache 2.0 you can use Qwen commercially, modify it, and distribute it, with the patent grant and change-notice requirements. That is the whole deal for Qwen license commercial use terms.
Here is one I see people miss: fine-tunes. If you are using a fine-tune of Qwen rather than the base weights, check both licenses. The fine-tune's terms layer on top of the base model's, and both apply to you. A license compliance audit SaaS can surface this kind of stacking, but honestly a checklist and ten minutes gets you there too.
Another layer: if you are using Qwen through a third-party host or API instead of self-hosting, that provider's terms layer on top as well. The model license covers the weights. The API terms cover the service. Read both. They can say different things, and the stricter one is the one that governs you.
When the paperwork is this clean, commercial use AI model clearance is mostly about documenting which checkpoint you shipped and which license it carried. Do that and move on. And run an LLM license comparison before you commit, because "Qwen" is a family name, not a license.
I track license terms per model version across 200 models at localaitools.fyi, and I write commercial-use clearance reports before you commit to a checkpoint.