Alibaba released Qwen3.6-Max-Preview on April 20, a proprietary hosted flagship model available via Qwen Studio and Alibaba Cloud Model Studio. The model ranked first across SWE-bench Pro, Terminal-Bench 2.0, SkillsBench, QwenClawBench, QwenWebBench, and SciCode, outperforming Claude 4.5 on several agentic coding tasks. It supports a 256K-token context window, accepts text input only at launch, carries no open weights, and ships with a novel preserve_thinking feature that maintains reasoning traces across multi-turn conversations—useful for autonomous coding agents. Alibaba described it as a preview with further improvements expected.