3 articles · Updated · The New York Times · Jul 20
Summary
Alibaba on Sunday released a preview of its newest Qwen AI model and said it would be open source, with performance described as comparable to Anthropic’s Fable.
Nearly 4% gains in Alibaba’s Hong Kong-traded shares followed the announcement, as investors bet Chinese open-weight models could pressure pricier closed systems from U.S. rivals.
Moonshot AI’s Kimi K3 has already intensified that pressure after posting benchmark results close to Fable and drawing such heavy demand that it was unavailable for much of the weekend.
Wednesday’s Alphabet earnings now loom as a fresh test of whether Big Tech can justify heavy AI spending while Chinese models narrow the performance gap at lower cost.
As China's AI advances despite chip bans, is the US strategy of controlling hardware already failing?
With divergent goals of economic integration versus AGI, which nation's AI strategy will ultimately dominate the future?
Qwen3.8-Max: Alibaba’s 2.4 Trillion-Parameter Model and the New Era of Open-Source AI Competition
Overview
Alibaba has introduced Qwen3.8-Max-Preview, a major new addition to its Qwen family of AI foundation models. With a massive 2.4 trillion parameters, Alibaba positions this model as one of the most powerful in the world, claiming it is second only to Anthropic’s Fable 5. This launch strengthens Alibaba’s presence in the global AI race and demonstrates its ambition to compete with leading frontier AI models. Qwen3.8-Max-Preview is now available through Alibaba’s Token Plan and Qoder platforms, marking a strategic move to expand its influence and accessibility in the rapidly evolving AI landscape.