Alibaba says its new model ranks second globally. It published no benchmark to back that up

Alibaba unveiled Qwen3.8-Max-Preview on July 19 at the World AI Conference in Shanghai, days after Moonshot AI's Kimi K3 open-weight launch — a 2.4-trillion-parameter mixture-of-experts model, the first Qwen system above one trillion parameters to handle text, images, video and documents in a single system. It carries a roughly 983,000-token context window, an output cap of 131,072 tokens, and an API endpoint built to accept requests formatted for both OpenAI's and Anthropic's schemas, so integrated teams don't have to rewrite their calling code to try it.

Alibaba's own material positions the preview as 'second only to' Anthropic's Claude Fable 5 among frontier systems, and says it beats its predecessor, Qwen3.7-Max, on coding, full-stack development, data analysis and office workflows. What's missing is anything a buyer could check: no score table, no task-level results, no model card and no independent evaluation accompanied the announcement. The ranking is Alibaba's word, not a reproducible result.

The preview is live now through Alibaba's Token Plan and Qoder platforms at roughly 10% of expected standard pricing; the outgoing Qwen3.7-Max charged $2.50 per million input tokens and $7.50 per million output tokens as a reference point, though Alibaba hasn't set final pricing for the new model. Full open weights are promised 'soon,' with no license, checkpoint or date attached yet. A 2.4-trillion-parameter count is a real number; a ranking with no published benchmark behind it is not — until Alibaba, or someone else, publishes one, 'second only to Fable 5' is a claim to note, not a result to cite.