Alibaba Unveils 'Qwen3.8-Max' Preview — 2.4 Trillion Parameters, Open Weights Promised
On July 19, 2026, Alibaba unveiled a preview of “Qwen3.8-Max,” the new flagship model in its Qwen lineup, via the Qwen team’s official X (formerly Twitter) account. With roughly 2.4 trillion parameters, Alibaba describes it as “one of the most powerful models currently available, on par with frontier-class models and second only to Fable 5.” The company also indicated it plans to release the model with open weights in the near future.
Details
- Parameter scale: 2.4 trillion parameters. Described as a model that is “still continuously evolving,” suggesting training is ongoing even now
- Performance claims: Alibaba claims it is “on par with frontier-class AI models, second only to Fable 5.” No detailed benchmark figures were disclosed in this announcement
- Availability: The preview is already accessible through Alibaba Cloud’s Token Plan and the coding platforms “Qoder” and “QoderWork”
- Promise of open weights: While currently offered as a closed preview, Alibaba has explicitly stated it plans to release the weights soon — a potential shift away from the “flagship models stay closed-API-only” approach that has held since the Qwen3.7 generation
- Where it fits in the series: The latest entry in a rapid string of flagship releases, following Qwen2.5-Max (January 2025), Qwen3 (April 2025), Qwen3.6, and Qwen3.7-Max (May 2026)
How to try it
- Developers can access the Qwen3.8-Max preview right now via Alibaba Cloud’s Token Plan, through the coding platforms “Qoder” and “QoderWork”
- Consumer availability and third-party benchmark verification have not yet taken place; further updates are expected, including on the timing of the open-weight release