Worried about Qwen 4 License
retained text
After Alibaba moved from Permissive OSS licenses to its own custom license with IMO ambiguous terms in models like Qwen3.8-Flash-Next, is there any indication this will carry forward into Qwen4? While that license at first reads permissive for local/internal use, fine-tuning and derivatives, it’s definitely not Apache/MIT. The two big catches are: 1. commercial MaaS or a standalone coding/office AI assistant requires a separate Qwen license seemingly from day one, and the wording around outputs is annoyingly vague. 2. The internal-use exception says you can’t make the model, its outputs, or capabilities available to third parties, but it **never clearly says whether downstream code/data produced indirectly from internal outputs is unrestricted.** The $20M/month or 100M-MAU threshold seems to be an attribution trigger, not the threshold for needing a commercial license. IMO output ownership needs to be clearly covered in the license, that wasn’t the case when I last checked. Edit: This is not legal advice but I think it’s fair to highlight. I don’t want to embrace the trend of moving away from tried and tested OSS licenses like apache and MIT. 3.8 Next was interpreted by many as a preview into 4 and thus might seem to imply a new licensing strategy. One could dream and hope the Alibaba team see that people care about this and **at least** add clearer licensing language.
document fetched 2026-10-30 00:45 · canonical https://reddit.com/r/LocalLLaMA/comments/1xifjn2/worried_about_qwen_4_license