Qwen3.6 Max Preview is the flagship proprietary language model from Alibaba's Qwen team, released in April 2026. As the highest-performing entry in the Qwen 3.6 series, it is designed to push the boundaries of agentic coding, world knowledge, and instruction following. The model is part of a broader product matrix that includes the Qwen3.6-Plus and Qwen3.6-Flash variants, as well as open-weight models like Qwen3.6-35B-A3B.
The model is built on an ultra-large-scale architecture, reportedly exceeding 1 trillion parameters. This scale enables significant improvements in complex, repository-level problem solving and real-world agent reliability. On major benchmarks such as SWE-bench Pro, Terminal-Bench 2.0, and SciCode, the preview release demonstrated substantial performance gains over its predecessors, achieving top scores in several agent-oriented evaluations.
A key technical feature of the Qwen 3.6 series is Thinking Preservation, which allows the model to retain reasoning context across multiple turns in a conversation. This capability is specifically optimized for iterative development and long-running agentic tasks where maintaining a coherent chain of thought is critical. The model supports a context window of 262,144 tokens, providing ample space for large codebases and extensive documentation.
While the Max tier remains proprietary and accessible via the Alibaba Cloud Model Studio API, it maintains compatibility with standard API protocols, including those from OpenAI and Anthropic. The preview release serves as an early look at the next-generation flagship, with Alibaba continuing to iterate on the model to enhance world knowledge and instruction-following precision before a full stable release.