Alibaba Opens Source Qwen 3.6-35B-A3B Mixture-of-Experts Model

Summary

Alibaba's Qwen team has released the Qwen 3.6-35B-A3B model as open source. The model name suggests a mixture-of-experts (MoE) architecture with 35 billion total parameters and approximately 3 billion active parameters per inference pass, making it a relatively efficient model for deployment. This follows Alibaba's ongoing strategy of releasing competitive open-weight models in the Qwen family to build ecosystem adoption. The release is notable in the context of the accelerating open-source mod