From the source
Alibaba's Qwen team released Qwen3.8-2.4T-A95B, the first open-weight Qwen-Max-class model with 2.4 trillion total parameters and 95 billion activated per token.
- The model features a hybrid linear-plus-full-attention architecture, native context up to 262K tokens extensible to 1M, and is designed for agentic and reasoning workloads.
Amazon published a guide on deploying this model on SageMaker HyperPod using vLLM on ml.p6-b300 instances.





