Alibaba published the Qwen 3.8 weights on Hugging Face in August: the 27B under Apache 2.0 and the 2.4T-A95B under a custom Qwen3.8-Max license.
The 27B has about 6.8 million downloads versus about 32,500 for the 2.4T, which needs roughly 2.45 TB of weights in FP8.
The Max license is MIT-like but needs a separate deal for model-as-a-service or AI work assistant businesses over $50 million in revenue.
The open 2.4T checkpoint is not the hosted Max product, which adds vision, non-thinking mode, 1M default context and built-in tools.