Alibaba launches flagship AI model to challenge Chinese and US rivals

Alibaba on Monday unveiled Qwen3.8-Max, its largest and most capable artificial intelligence model to date, as the Chinese technology giant seeks to strengthen its position in the global AI race with a system that rivals leading domestic and international competitors.

According to the company, Qwen3.8-Max contains 2.4 trillion parameters, approaching the 2.8 trillion used by Moonshot AI’s Kimi K3 model released last month. While parameter count does not directly determine model quality, it is widely regarded as an indicator of the scale of computing resources and training data behind advanced AI systems. The model will be released next week through Alibaba Cloud’s Model Studio platform as an open-weight system, allowing developers to download and adapt it.

Alibaba said Qwen3.8-Max immediately became the highest-ranked Chinese text model on the Arena.AI benchmarking platform, although it remains behind Anthropic’s Claude Fable 5 and three Opus variants globally.

On Arena.AI’s leaderboard for image analysis, the model ranked second worldwide behind a Claude Fable 5 variant. Both Qwen3.8-Max and Kimi K3 support text, image and video inputs and can process up to one million tokens in a single context window.

Alibaba said the model uses a mixture-of-experts architecture that activates only 95 billion of its 2.4 trillion parameters for each task, reducing computational costs and response times. The company added that the system completed a software engineering project over 16 days during internal testing, highlighting its ability to perform extended autonomous tasks.

Alibaba said one internal evaluation saw Qwen3.8-Max work independently for around five days to reproduce and improve a mathematics research experiment, while another test placed it ahead of all but 68 of 526 human teams in a 24-hour programming competition. These results have not been independently verified and that Anthropic has made similar claims for its latest Claude models, which are designed to complete complex coding and research tasks with limited human oversight.

Alibaba’s latest release follows a series of high-profile launches by Chinese AI developers including Moonshot, DeepSeek and ByteDance, reflecting intensifying competition as Chinese firms attempt to close the gap with US leaders despite restrictions on access to advanced semiconductor technology.