Alibaba Group on Monday, February 16 unveiled Qwen3.5, a new flagship artificial intelligence model designed to execute complex tasks independently, in a move the Chinese tech giant says ushers in an “agentic AI era” and raises the stakes in the global race against U.S. rivals.
Cheaper, more powerful ‘agentic’ model
Alibaba said Qwen3.5 is 60% cheaper to use and up to eight times more capable of handling large-scale workloads than its immediate predecessor, significantly lowering the cost of deploying advanced AI for businesses and developers. The company positions the model as a new benchmark for “capability per unit of inference cost”, arguing that users can now run more sophisticated AI applications without proportionally higher compute bills.
Built for what Alibaba describes as the “agentic AI era”, Qwen3.5 is engineered not just to respond to prompts but to take actions across applications, autonomously planning and executing multi-step tasks. The model features so‑called visual agentic capabilities, allowing it to understand on‑screen interfaces and operate across both mobile and desktop apps, effectively turning it into a digital operator that can navigate services on behalf of users.
Strategic push behind Qwen chatbot
The launch is central to Alibaba’s push to grow the Qwen chatbot app, which sits at the heart of the group’s broader consumer and enterprise AI strategy. In China’s increasingly crowded AI market, Qwen currently trails ByteDance’s Doubao and rising star DeepSeek, which last year became the first Chinese AI firm to achieve global breakout status.
Alibaba is using aggressive growth tactics around Qwen, including a recent campaign of in‑app coupons tied to food and beverage purchases that drove a seven‑fold increase in active users despite technical glitches. The company has also been integrating Qwen more deeply into its ecosystem, from e‑commerce and payments to travel and mapping services, allowing users to browse, compare and complete purchases through a single AI interface.
Competition with ByteDance, DeepSeek and U.S. rivals
The unveiling of Qwen3.5 comes just days after ByteDance rolled out Doubao 2.0, an upgraded model that powers one of China’s most widely used chatbot apps, with close to 200 million users. ByteDance has also framed its latest model as built for the age of AI agents, underscoring how leading Chinese platforms are converging on autonomous task execution as the next battleground.
DeepSeek, which shook up the industry last year with a highly capable, low‑cost model that gained traction well beyond China, is also expected to bring new iterations to market, further intensifying competition at home and abroad. Industry analysts say Qwen3.5’s efficiency claims and agentic focus could help Alibaba close the gap with domestic rivals while positioning its stack more competitively against top‑tier U.S. models in tasks that blend reasoning, planning and real‑world execution.
Alibaba has previously said its Qwen series performs on par with leading international systems on a range of benchmarks, especially in reasoning‑heavy and coding scenarios, as part of a broader push to make open and hybrid models a core pillar of its cloud strategy. The company is also leaning on open‑source dynamics around earlier Qwen generations, which have logged hundreds of millions of downloads and spawned tens of thousands of derivative models, to accelerate adoption of its latest releases.
From ‘AI that responds’ to ‘AI that acts’
Qwen3.5 builds on a year‑long shift inside Alibaba from conversational assistants toward operational agents that can orchestrate services across the group’s sprawling platforms. In the upgraded Qwen app, the model can, for example, search Taobao for products, compare prices, apply coupons, complete payments through Alipay and arrange delivery or travel bookings via Fliggy and Amap, with minimal user intervention.
Alibaba describes this as moving users from intent to completion inside a unified AI layer, with Qwen3.5 acting as the logic engine that plans, coordinates and verifies each step. For enterprise customers, the company says the same agentic capabilities can be used to automate workflows such as customer support, internal knowledge retrieval, supply‑chain coordination and software operations, with the model able to call tools and interact with internal systems under predefined security constraints.
Implications for Alibaba and global AI
The launch of Qwen3.5 comes as Chinese tech firms race to narrow the perceived gap with leading U.S. players, while also managing rising regulatory and geopolitical pressures around advanced AI exports. By emphasizing lower cost, higher throughput and practical agent‑like functions over purely headline benchmark scores, Alibaba is signaling that commercial viability and day‑to‑day automation are now as important as raw model performance.
Investors and analysts will be watching how quickly Qwen3.5 is adopted across Alibaba Cloud clients and the wider developer community, and whether its claimed efficiency edge translates into revenue growth and user gains in the months ahead. With rival models from ByteDance, DeepSeek and U.S. providers also evolving rapidly, Qwen3.5 marks both a milestone in Alibaba’s AI roadmap and the start of a new phase of competition focused squarely on the emerging agentic AI era.
Post Comments
Be the first to post comment!
Related Articles
