Alibaba's Qwen team has released Qwen3.8-Max, a 2.4-trillion-parameter model it describes as the most capable in the Qwen family to date.
According to MarkTechPost, the model has moved from preview to general availability with published per-token pricing. It uses a mixture-of-experts design — meaning only a fraction of those 2.4 trillion parameters are active for any given request — and accepts text, image and video input across a context window of one million tokens, roughly the amount of material you could feed it in a single go.
The competitive framing is unusually direct. Euronews reports that Alibaba's new AI "claims to match Claude," Anthropic's flagship, while Yahoo Tech frames the launch as a challenge to both OpenAI and Anthropic. Bloomberg's Luz Ding, via Techmeme, reports that Alibaba says Qwen3.8-Max tops Kimi K3 on some benchmarks, and that the company plans to release open weights for both Qwen3.8-Max and a smaller Qwen3.8-27B next week.
Those performance claims deserve a caveat. MarkTechPost notes that no benchmark table has been published, so for now the comparisons rest on the company's own statements rather than independently verifiable numbers. The promised open-weights release next week would change that, letting outside researchers test the claims themselves.
Developers are paying attention: Alibaba's announcement post, titled "A New Bar for Coding and Cowork," reached the front page of Hacker News with 107 points and 36 comments.
Why it matters: if a Chinese lab can ship a frontier-class model and then give away the weights — something the leading US labs largely don't do with their top models — the competitive advantage of America's best-funded AI companies starts to look a lot less durable.