Qwen3.8-Max Tool Calling: A Python Agent Loop

On August 3, 2026, Alibaba's Qwen team shipped Qwen3.8-Max, a 2.4-trillion-parameter mixture-of-experts model that activates about 95 billion parameters per token. It takes text, images, and video, carries a one-million-token context window, and was built for long-horizon agentic work: running multi-step tasks on its own for hours. On the agentic benchmarks people actually argue about it is right in the mix with the frontier, scoring 86.6 on Terminal-Bench 2.1 and topping PaperBench at 93.0. That is why it is the model everyone is wiring into their agents this week.