AI technology

Alibaba launches Qwen4-Max; smaller open-weight models remain promised

Alibaba's November 16 announcement describes hosted access to Qwen4-Max and promises smaller open-weight Qwen4 models later. Its capability claims remain company-reported.

Event date: Nov 16, 2026

What the announcement establishes

Alibaba announced Qwen4-Max on November 16, describing native vision and a one-million-token context window. It reports a mixture-of-experts architecture with 2.4 trillion total parameters and 71 billion activated per token. [1]

Spicy's takeThe release quotes $0.075 per million input tokens for implicit cache hits. That is a specific cached-input rate, not a complete price for an arbitrary request; readers should check current ordinary-input and output charges before budgeting. [1]

Access and evidence are separate questions

Spicy's takeAnalysis: hosted availability and downloadable weights are different milestones. For readers choosing a local deployment, the smaller-model promise is something to watch, rather than a usable release established by this announcement. [1]

Spicy's takeAlibaba presents benchmark and workflow gains as reasons to adopt the model. This report has not independently tested those claims or account access, and does not establish better real-world performance. The useful confirmed distinction is what Alibaba announced now versus what it says will follow. [1]

Sources & further reading

Source material checked Nov 17, 2026. Linked material may change after publication.

  1. Alibaba unveils Qwen4-Max โ†—

    Alibaba Cloud ยท Primary source ยท Nov 16, 2026