Alibaba has introduced its largest AI model to date, Qwen3.8‑Max, boasting 2.4 trillion parameters. While slightly smaller than Moonshot AI’s Kimi K3 (2.8 trillion), the model’s mixture‑of‑experts architecture promises lower costs and faster responses.

Key Takeaways

  • Alibaba’s Qwen3.8‑Max features 2.4 trillion parameters.
  • It trails Moonshot’s Kimi K3 (2.8 trillion) by just 0.4 trillion.
  • The mixture‑of‑experts design activates only 95 billion parameters per request, cutting costs.

On Monday, Alibaba officially unveiled Qwen3.8‑Max, its most capable AI model to date. With 2.4 trillion parameters, the model places China’s AI race a step ahead, yet remains only marginally smaller than Moonshot AI’s recently launched Kimi K3.

Historical Background

Over the past decade, Chinese tech giants have accelerated the development of open‑weight AI models, seeking to capture global market share while reducing reliance on foreign hardware. Companies like Alibaba, Baidu, and Huawei have repeatedly pushed the parameter frontier to demonstrate computational prowess and attract developer ecosystems.

ModelParameters (trillion)Capabilities
Qwen3.8‑Max (Alibaba)2.4Text, Image, Video; 1 million token limit
Kimi K3 (Moonshot AI)2.8Text, Image, Video; 1 million token limit

Why This Matters

BozokMedia analysis shows that publishing parameter counts not only fuels developer hype but also signals a commitment to open‑source collaboration. Alibaba’s mixture‑of‑experts architecture—activating just 95 billion parameters per query—delivers a cost‑effective edge in the fiercely competitive AI landscape.

"Qwen3.8‑Max’s expert‑mix design makes large‑scale AI both affordable and scalable," notes AI researcher Dr. Li Wen.
Did You Know?: Qwen3.8‑Max completed a software‑engineering project in just 16 days, roughly half the industry‑standard timeline.

Frequently Asked Questions

Q1: What tasks can Qwen3.8‑Max handle?
It processes text, images and video, and can ingest up to 1 million tokens in a single request.

Q2: Is the model open for developers?
Yes, Alibaba released it as an open‑weight model, allowing developers to download and fine‑tune the underlying parameters.