Alibaba has unveiled Qwen3.8-Max, a new artificial intelligence model the Chinese company says rivals the performance of leading closed-source systems despite its open structure.
According to Alibaba, Qwen3.8-Max contains a total of 2.4 trillion parameters. The model does not use all of them at once, instead relying on a Mixture of Experts architecture that activates roughly 95 billion parameters at a time, a design Alibaba says lowers processing costs and reduces latency while still allowing for much greater capacity.
Alibaba returns to open source
The release also marks a shift in strategy for Alibaba, which had offered some of its top-tier models as closed source earlier this year. With Qwen3.8-Max, the company is returning to an open-source approach. Alibaba said it would publish the model's weights next week so developers can download and use them, making Qwen3.8-Max the first Max-class Qwen model to be released as open source.
1 million token context window
Alibaba said one of the model's standout features is a context window of 1 million tokens, letting it analyze hundreds of pages of documents, long video recordings or extensive datasets in a single pass. The company also said the model can build searchable knowledge bases from video content and livestreams.
Alibaba is positioning Qwen3.8-Max for autonomous software development, long-term research projects and AI agents that need to run continuously for days. In the company's internal testing, the model independently carried out and completed a software engineering project over 16 days.
