Alibaba Previews Qwen3.8-Max, a 2.4 Trillion-Parameter Multimodal Model, Days After Moonshot’s Kimi K3 Open-Weight Launch

Editor's take

Alibaba has unveiled a preview of Qwen3.8-Max-Preview, a massive 2.4 trillion-parameter Mixture-of-Experts (MoE) multimodal model, positioning it as a significant step in large model development.

This release, coming shortly after Moonshot AI's Kimi K3 open-weight model, intensifies the competition in the Chinese AI landscape, particularly for high-performance, multimodal capabilities. The sheer parameter count suggests an ambition to rival top-tier models, though its practical impact remains to be seen without performance benchmarks.

The immediate focus will be on the availability and performance of Qwen3.8-Max-Preview in real-world applications, especially as it's offered at a reduced price. Key questions revolve around its actual efficiency compared to its stated parameter count and how it stacks up against competitors like Fable 5 and the already accessible Kimi K3 in practical tasks.