Alibaba Group Holding opened a preview of Qwen4-Max, the flagship model in its next-generation Qwen family, on November 111,2. The multimodal model accepts image and video inputs alongside text and supports a context window of one million tokens.
Alibaba says Qwen4 brings deeper knowledge understanding, richer perception of the world, and stronger coding and agentic abilities.
Qwen4-Max-Preview is available only through Alibaba's TokenPlan subscription, not pay-as-you-go. Alibaba says calls consume TokenPlan credits at a promotional rate as low as 20 percent of the standard rate. The preview runs from 10:00 Beijing time on November 11 to midnight at the end of November 25, and the documentation states the model will not change during that window and will be retired afterward.
Output is capped at 393,216 tokens, shared between reasoning and the final answer. Thinking is on by default and can be adjusted with a reasoning-effort setting. The model supports web search, function calling, built-in tools, and structured JSON outputs. It is compatible with OpenAI's Chat Completions and Responses APIs and Anthropic's Messages API.
Alibaba Cloud's Model Studio documentation describes Qwen4-Max-Preview as targeting coding, office work, front-end design, and complex agentic tasks. Early access is also available in QwenWork, the Qoder coding tool, the Qwen app, and QwenStudio.
ANALYSIS The two-week, subscription-only preview with no published benchmarks or weights positions Qwen4-Max as a controlled stress test rather than a general release. The 393,216-token output cap, split between chain-of-thought reasoning and the final answer, is a notable design choice, though its competitive standing depends on benchmarks Alibaba has yet to provide.
Alibaba's chip design unit T-Head separately filed confidentially for a Hong Kong initial public offering, advancing a spin-off tied to the group's AI and semiconductor assets[3].