The Drum · Weekly AI news, freeSubscribe →
Share
Qwen3.5 introduces a groundbreaking hybrid architecture that combines linear attention and sparse mixture-of-experts, making it highly efficient and capable of advanced multimodal tasks.
We're excited to announce the official release of Qwen3.5, particularly the open-weight model Qwen3.5-397B-A17B. This new version marks a significant step forward in multimodal AI, offering impressive capabilities across reasoning, coding, agent functionalities, and multimodal understanding. Here's what you need to know:
Qwen3.5-397B-A17B was evaluated against leading models in various tasks, demonstrating competitive and often superior performance:
For those who prefer a hosted solution, Qwen3.5-Plus is available via Alibaba Cloud Model Studio:
For practitioners, Qwen3.5 represents a significant advancement in multimodal AI. The combination of linear attention and sparse MoE ensures that the model is both powerful and efficient, making it suitable for real-world applications. Whether you're working on language understanding, coding tasks, or building complex agents, Qwen3.5 offers a robust foundation.
You can try out Qwen3.5 through various platforms:
Tags
Original Sources
About the author
Kai built ML infrastructure at a Bay Area startup before developing an obsession with transformer architectures and inference optimisation that eventually pulled him out of product work entirely. A stint at a compute research lab sharpened his instinct for what actually matters in a model release versus what is marketing. He writes from the inside — from the perspective of someone who has debugged the systems he is describing at three in the morning. He is allergic to hype and instinctively drawn to the unglamorous plumbing questions that everyone else skips over.
More from The Engineer →This Week's Edition
17 February 2026
88 articles
Related Articles
Related Articles
More Stories
