Alibaba unveils 2.4T parameter Qwen3.8, open-weight release coming

Alibaba just dropped word on its next flagship AI model, and the parameter count is eye-watering.

The Qwen team announced Qwen3.8, a massive 2.4 trillion parameter model that will soon be released as open weights. According to Alibaba, it is currently one of the most capable models in the world, ranking second only to Fable 5 among frontier AI models.

A preview is already live

While the full open weights are not out yet, you do not have to wait to test it. Alibaba rolled out a preview version, Qwen3.8-Max-Preview, on its Token Plan platform.

It is also accessible inside Alibaba's developer and workspace tools, Qoder and QoderWork, allowing early testing ahead of the public weight drop.

Why it matters

We think releasing open weights at a 2.4T scale is a massive statement.

Frontier models of this size have almost exclusively lived behind proprietary, closed-source APIs. If Alibaba delivers on its promise, researchers and enterprises will get their hands on true frontier-tier weights without being locked into closed ecosystems.

That said, running a 2.4 trillion parameter model locally is completely out of reach for everyday developers. For now, most teams will still interact with it through cloud interfaces, and the claim of being second only to Fable 5 still awaits independent community benchmarks.