Alibaba Group

AI · Cloud Computing · China Tech

Launch

Alibaba launches Qwen3.8-Max, 2.4-trillion-parameter model rivaling Claude

Open-weight releases like this let China's AI labs turn transparency into a competitive edge over secretive U.S. rivals.

• Alibaba unveiled Qwen3.8-Max on Monday, calling it its largest and most capable AI model yet, with the model's weights set to be released next week through Alibaba Cloud's Model Studio platform.
• Qwen3.8-Max extends Alibaba Cloud's open-weight Qwen model family, which the company folds directly into its own cloud, e-commerce and consumer services to give each new model an immediate base of users.
• The model runs on a 2.4-trillion-parameter mixture-of-experts architecture but activates only about 95 billion parameters per query, a design meant to keep a huge model cheap and fast to run.
• Qwen3.8-Max supports a 1-million-token context window, enough to process a large codebase, a roughly 100-hour video, or hundreds of pages of documents in a single prompt.
• In an internal test, the model autonomously built a command-line coding tool over 16 days, logging 265 commits, 127 pull requests and 151 issues without any human touch.
• Alibaba's New York-listed shares climbed about 4.5-4.8% and its Hong Kong shares jumped 7% the day the model was announced.
• The release keeps Alibaba in close contention with Moonshot AI's Kimi K3, a 2.8-trillion-parameter model that had briefly stood as China's largest open AI system.
• By publishing exact parameter counts and open weights while OpenAI and Anthropic keep such details private, Chinese AI labs are making transparency itself a competitive differentiator in the global AI race.