According to ChainCatcher, Alibaba released Qwen4-Max on Nov. 16. The model features 2.4 trillion total parameters with 71 billion active parameters per token, native vision capabilities, and a 1 million token context window. Qwen4-Max ranks sixth on Code Arena and achieved top scores across multiple benchmarks including Agents' Last Exam and JobBench.
In Alibaba's Work360 benchmark, Qwen4-Max outperformed its predecessor Qwen3.8-Max, reducing token consumption by 35% and cutting average task costs by approximately 60% in high-intensity reasoning mode. The model is available via QwenCloud's Token Plan and API at $0.075 per million input tokens.