Overview
- Alibaba has launched Qwen3.8-Max and says it competes with leading U.S. systems such as OpenAI and Anthropic.
- The company reports the model has 2.4 trillion parameters and calls it its largest system, highlighting strengths in programming, long-form research and complex task handling.
- Arena.AI benchmark results place Qwen3.8-Max as the highest-ranked Chinese model on text tasks while showing it trails some U.S. models overall and scores especially well on image evaluation.
- Qwen3.8-Max is available for immediate use and Alibaba plans to publish the model weights next week so developers can run, adapt and fine-tune the system themselves.
- The release follows a rapid wave of high-parameter Chinese models, including Moonshot AI’s 2.8 trillion-parameter Kimi K3, and signals growing competition that could speed local developer adoption and increase pressure on U.S. providers.