Qwen3.8-Max launches through QwenCloud

Alibaba's Qwen team introduced Qwen3.8-Max on August 2, 2026. The model became available through QwenCloud, with the release focused on coding, research, and multi-step work.

A larger Max model

Qwen3.8-Max used 2.4 trillion total parameters and 95 billion active parameters. The announcement described training across tasks involving code, files, tools, and longer-running workflows. It also included evaluations of the model in several agent frameworks.

Hosted access before the weight release

API access was available through QwenCloud on announcement day. The team planned to publish the model weights on Hugging Face and ModelScope the following week. That was the first planned downloadable release of a Qwen-Max-class model.

The launch opened hosted access immediately, while self-managed deployment depended on the separate weight release.