Qwen3.6-Max-Preview

Top proprietary tier of the Qwen3.6 generation and the first fully closed-weights Qwen Max flagship — released only as a Preview ("smarter, sharper, still evolving"); no GA Qwen3.6-Max ever shipped. 256K context, text-only, with a preserve_thinking mode recommended for agentic work and an API compatible with both OpenAI and Anthropic specs. $1.30/$7.80 per M tokens (90% cache-hit discount).

Alibaba claimed #1 on six coding/agentic benchmarks (SWE-bench Pro, Terminal-Bench 2.0, SkillsBench, QwenClawBench, QwenWebBench, SciCode), with gains over Qwen3.6-Plus of +9.9 SkillsBench and +10.8 SciCode. AA Intelligence Index v4.1: 40. Parameters undisclosed (press reports of "35B-A3B" conflate it with the open Qwen3.6-35B-A3B). Deprecated a month later in favor of Qwen3.7-Max (API live May 19, 2026).