Alibaba Cloud Qwen4-Max Preview Launches on Qwen AI and Bailian Platform, Token Plan Opens Simultaneously
Alibaba Cloud launched Qwen4-Max (model identifier qwen4-max-preview) as a preview on November 11, 2026, with API services and the Token Plan opening for calls simultaneously; the Lite version early-access price is 19.50 yuan/1 month. The preview window runs from November 11 at 10:00 to November 25 at 24:00 (UTC+8); after the window closes, the identifier will be taken offline and the official version will be released under a new name. The company says it uses a MoE architecture based on Qwen Sparse Attention (QSA), supports a 1M Token context and text, image, and video inputs, with coding and Agentic as the focus of this upgrade; on Terminal-Bench 4.0, the average Token usage per attempt dropped from 289K to 151K, and the score rose from 37.9 to 48.5. On the China site, input is 10 yuan per million Tokens, output 30 yuan, and implicit cache hits 0.5 yuan.
Why read this
SelectedAI score 85
The original text provides the preview window, pricing, and Terminal-Bench 4.0 comparison data, which can be used to assess integration costs and task suitability.
AI Trends