Alibaba

AI Models · Cloud Computing

Launch

Alibaba launches 2.4T-parameter Qwen3.8, claims near-frontier performance

The preview is priced at a steep discount, turning the model into a funnel that locks developers into Alibaba's cloud ecosystem.

• Alibaba's Qwen team unveiled Qwen3.8-Max-Preview, a 2.4-trillion-parameter model, during the World AI Conference in Shanghai, calling it 'second only' to Anthropic's Claude Fable 5.
• Qwen3.8 is built on a sparse Mixture-of-Experts architecture and is the team's first multimodal model to exceed one trillion parameters, handling text, images, video and documents.
• The preview is live now on Alibaba's Token Plan, Qoder and QoderWork at just 10% of standard pricing, with a full open-weight release promised 'soon.'
• At 2.4 trillion parameters, Qwen3.8 is the second-largest publicly disclosed model, trailing only Moonshot AI's 2.8T-parameter Kimi K3 — a lab in which Alibaba holds a roughly 36% stake.
• Alibaba has not published independent benchmarks backing its 'second only to Fable 5' claim, nor disclosed how many of the 2.4 trillion parameters are active per query.
• The preview arrives just days before Kimi K3 goes fully open-source on July 27, timing that suggests Alibaba is racing to pull developers into its own stack first.
• The parameter race among Chinese AI labs shows compute-heavy scaling is outpacing independent verification, with companies prioritizing ecosystem lock-in over confirmed capability claims.