Alibaba’s Qwen 3.8-Max, built for big tasks including text, coding, and multi-step long horizon agentic capabilities, scales to 2.4 trillion parameters. Alibaba plans to release the model weights so companies can run it on-premise and on-device, a move shaking up the usual closed approach of US tech giants. Meanwhile, as AI costs skyrocket for companies like Alphabet, Meta, and Microsoft, efficient models like this could be a game changer for keeping budgets in check.