Alibaba Group has introduced a new artificial intelligence model, Qwen3.5, built to handle complex tasks on its own. The company says the model delivers major gains in both performance and cost efficiency, outperforming leading U.S. competitors across several benchmarks.
The launch is part of Alibaba’s push to grow its Qwen chatbot app in China’s crowded AI market. That space is currently led by ByteDance’s Doubao and rising star DeepSeek, which became the first Chinese AI company to achieve a significant global breakthrough last year.
Alibaba says Qwen3.5 costs 60% less to run than the previous version and processes large workloads eight times more efficiently. The model also introduces what the company calls “visual agentic capabilities,” meaning it can independently carry out actions across mobile and desktop apps. This marks a shift toward AI systems that do more than respond to prompts—they can complete tasks on behalf of users.
In a company statement, Alibaba described Qwen3.5 as built for the “agentic AI era,” aimed at helping developers and businesses move faster while getting more output from the same computing power. The company framed the release as a new benchmark for capability relative to inference cost.
Just days earlier, ByteDance rolled out Doubao 2.0, an upgraded version of its chatbot app, which now has close to 200 million users in China. Like Alibaba, ByteDance positioned its update as part of the move toward AI agents capable of independent action.
Qwen3.5’s debut could strengthen Alibaba’s recent momentum in China’s fiercely competitive AI model race. Earlier this month, the company ran a coupon campaign that encouraged users to order food and drinks directly through the Qwen chatbot. The promotion drove a sevenfold jump in active users, though some technical glitches were reported.
Last year, Alibaba moved quickly after DeepSeek’s sudden surge in popularity, releasing Qwen 2.5-Max and claiming it outperformed one of DeepSeek’s most talked-about models. In the announcement for Qwen3.5, however, Alibaba did not directly reference DeepSeek. The benchmarks it shared compared the new model to its predecessor and to major U.S. systems, including GPT-5.2, Claude Opus 4.5, and Gemini 3 Pro.
DeepSeek is expected to unveil its next-generation model in the coming days. That release is drawing close attention from investors and industry observers, especially after the company’s rapid rise last year triggered a sharp selloff in global technology stocks.
