Qwen 3.8-Max launches with a focus on extended agent work
The August announcement scales the flagship to 2.4 trillion total parameters and promises open weights.
THE STORY IN THREE POINTS
- The official launch article is dated August 3.
- The model has 2.4 trillion total and 95 billion active parameters.
- The launch promised open weights for the following week.
The announcement
Qwen announced 3.8-Max on August 3 as a 2.4-trillion-parameter model with 95 billion active parameters, available through QwenCloud. The launch emphasizes coding, knowledge work and tasks sustained over long periods. It also promised the first open weights for a Qwen-Max-class model in the following week. That is a dated launch commitment; this brief does not independently confirm the subsequent delivery of those weights.
Editorial context
A long autonomous demonstration can show persistence while leaving the amount of human preparation unclear. For readers, the informative details are the initial instruction, available tools, execution limits and criteria used to accept the final result. Preserving those conditions makes it easier to distinguish a repeatable capability from an unusually successful example selected for a launch.
Editorial context: what to watch
A useful trial could repeat the same multi-stage task several times and record intervention, review effort and final quality. Larger parameter counts alone do not settle whether an application should switch models. The comparison should include the full cost of finishing work and identify which parts of the advertised workflow remain dependent on the surrounding agent tools.
Sources: Qwen: 3.8-Max official launch article, August 3 ↗
