OPEN LINK ↗
// 25m agoMODEL RELEASE
Alibaba drops Qwen4-Max-Preview with 1M context
Alibaba Cloud debuted a preview of its next-generation flagship multimodal model on Model Studio through November 25. The model combines text, image, and video comprehension with deep reasoning and a 393k output ceiling across a 1-million-token context window.
// ANALYSIS
Alibaba is signaling that Qwen 4 aims directly at frontier reasoning models from OpenAI and Anthropic, though gating the preview behind proprietary Token Plans contrasts sharply with Qwen's open-weights playbook.
- –Massive 393k output limit dynamically shares tokens between internal chain-of-thought reasoning and the final response
- –Native OpenAI and Anthropic API compatibility makes it a drop-in evaluation target for agentic tools like Cursor and Claude Code
- –Multimodal input support spans text, image, and video inputs out of the box
- –Strict two-week availability window without weights suggests Alibaba is testing compute and safety boundaries before any open release
// TAGS
qwen4-max-previewqwenllmmultimodalreasoninglong-contexttool-use
DISCOVERED
25m ago
2026-11-14
PUBLISHED
51m ago
2026-11-14
RELEVANCE
9/ 10
AUTHOR
GlobalFeedAI