Alibaba introduced Qwen3.8-Max as an upgrade to the Qwen3.5 generation. The company describes improvements in coding, office productivity, research, and tasks that require sustained work over long periods.
The launch announcement reported 2.4 trillion total parameters and planned to release the weights the following week. Total parameter count is not the same as the amount of computation used for every token, and it does not by itself determine practical quality.
The multimodal examples include analyzing hundreds of PDF pages and locating details in very long video collections. Qwen also describes extended repository work involving coding, tests, and project maintenance, as well as business-analysis workflows that adapt to new data.
These are the company's demonstrations and claims. For any large project, I would still check whether the model preserves details and verifies its changes. Read the official Qwen blog for the release information.
Adapted from the original Chinese article, published on August 3, 2026.
Comments NOTHING