Alibaba Group Holding (NYSE:BABA) has launched its most advanced artificial intelligence model to date, Qwen 3.5, claiming the system outperforms major U.S. competitors while dramatically lowering the cost of high-level machine reasoning.
The rollout is a direct strike at the "agentic" AI era, where models no longer just generate text but independently execute multi-step tasks across desktop and mobile applications.
Alibaba stated that Qwen 3.5 is 60% cheaper to operate and eight times more efficient at processing massive workloads than its predecessor, Qwen 3.0.
The release intensifies a brutal arms race in China’s AI sector.
Alibaba’s announcement comes just 48 hours after ByteDance launched Doubao 2.0, an upgrade to China’s most popular chatbot, which similarly prioritizes autonomous agent capabilities.
Alibaba is banking on "visual agentic" features—the ability for the AI to "see" and interact with app interfaces—to attract developers.
To spur adoption, the company recently executed a 3-billion-yuan ($400 million) coupon campaign, encouraging users to shop directly via the Qwen chatbot, which reportedly led to a seven-fold increase in active users.
In a bold performance claim, Alibaba published benchmarks suggesting Qwen 3.5 surpasses leading American models, including OpenAI’s GPT-5.2, Anthropic’s Claude Opus 4.5, and Google’s Gemini 3 Pro, in specific categories of reasoning and cost-per-inference.