Alibaba has unveiled its most capable artificial intelligence model to date, a 2.4 trillion-parameter system that the Chinese technology giant says rivals the best offerings from OpenAI and Anthropic at a fraction of the cost.
The Hangzhou-based company launched Qwen3.8-Max on August 3, 2026, calling it the largest and most sophisticated model ever produced by a Chinese lab. With roughly 95 billion active parameters at any given time for efficiency, the system ranks near the top of combined text and vision leaderboards, placing it within striking distance of Anthropic's most advanced systems according to benchmark results published alongside the release.
What distinguishes Qwen3.8-Max from its American competitors is not only its scale but its pricing. Alibaba set input costs at approximately two dollars per million tokens and output at six dollars per million tokens, a rate that dramatically undercuts comparable U.S. models. The company also committed to open-sourcing the model's weights, a strategy that has become central to China's push for broad developer adoption worldwide.
In autonomous testing, the model completed a full software engineering project over sixteen days without human intervention, demonstrating capabilities that Alibaba says place it firmly in the top tier of systems capable of extended reasoning and coding tasks. The announcement sent Alibaba shares climbing as much as seven percent in Hong Kong trading, reflecting investor enthusiasm for the company's intensified bet on generative AI.
The release arrives amid a fierce price war in AI inference, with Chinese labs including DeepSeek racing to offer powerful open-weight alternatives to closed U.S. systems. For Alibaba, which has poured billions into cloud infrastructure and AI research, Qwen3.8-Max represents both a technical statement and a commercial gambit aimed at capturing global developer mindshare in a market where cost increasingly matters as much as capability.
Image source: i.ibb.co