Alibaba has launched a preview of its new flagship AI model, Qwen3.8 Max, and is making a bold claim about where it stands among the world's best.
According to Bloomberg (via Techmeme), the model carries 2.4 trillion parameters — a rough measure of an AI system's scale — and Alibaba describes it as comparable to frontier AI models, ranking second only to Anthropic's Fable 5. The South China Morning Post similarly reports that Alibaba says the model's strength trails only Fable 5.
Perhaps the most significant detail is Alibaba's plan for how it will share the technology. Bloomberg reports that the company intends to make Qwen3.8 "open-weight soon." An open-weight release means the trained model's underlying parameters are made publicly available, letting developers and companies download and run it themselves rather than only accessing it through a paid service. That approach contrasts with the more closed model favored by some leading Western AI labs.
The preview was announced on Alibaba's Qwen account on social platform X, and it drew attention on Hacker News, where posts about the launch and its open-weight plans each gathered more than 100 points and dozens of comments. A related Qwen Cloud pricing page also circulated on the site.
Crypto Briefing framed the release as Alibaba "challenging Anthropic's dominance," underscoring how the announcement is being read as a competitive move against the current frontier leaders.
It's worth noting that the performance claims come from Alibaba itself; the sources here report the company's own characterization of the model rather than independent benchmark verification.
Why it matters: If a Chinese tech giant is releasing a top-tier AI model as open-weight, it could put frontier-level capabilities directly into the hands of developers worldwide, intensifying both global competition and the debate over how openly powerful AI should be shared.