Alibaba has released the weights for Qwen3.8-2.4T-A95B — for the first time, a model at the Qwen-Max level is being published as open source.
Without diving too deep into technical details, this puts it in the same league as top-tier Claude and GPT.
And that’s a big deal. Until now, the most powerful models were always closed, and open-source alternatives were noticeably weaker. But the situation is changing: now any company (if they can afford the compute) can self-host a truly strong model for their own use cases — and no longer needs to pay for a
Claude subscription or API usage.
Which raises a question:
Does this undermine the business model of the major AI API providers?
Screenshot huggingface
Top comments (0)