Alibaba unveiled Qwen3.5 on Feb 16, describing the model as built for what it calls the "agentic AI era" - systems that can act independently to complete complex tasks. The company highlighted two headline technical claims: the model is 60% cheaper to operate than its immediate predecessor, and it processes large workloads eight times more effectively than that prior iteration.
Beyond raw performance and cost metrics, Alibaba said Qwen3.5 includes "visual agentic capabilities," enabling the model to take autonomous actions across mobile and desktop applications. In a company statement, Alibaba framed the release as a means to help developers and enterprises "move faster and do more with the same compute, setting a new benchmark for capability per unit of inference cost."
Alibaba published a set of benchmarks alongside the announcement. Those comparative results show Qwen3.5 outperforming the company’s previous version and selected U.S. models cited by name in the materials: GPT-5.2, Claude Opus 4.5, and Gemini 3 Pro.
The launch arrives as Alibaba seeks to expand adoption of its Qwen chatbot app in China, where competition is intense. ByteDance recently upgraded its chatbot to Doubao 2.0, positioning that release similarly for the AI agent era. ByteDance’s app currently holds the largest user base in China for chatbots, approaching 200 million users.
Alibaba has already taken several steps to increase engagement around Qwen. Earlier this month the company ran a coupon giveaway that incentivized purchases of food and drink through the Qwen chatbot, an initiative that the firm says drove a seven-fold increase in active users despite encountering some glitches during the promotion.
The company’s Qwen3.5 announcement did not reference DeepSeek, one of the domestic startups that rose rapidly in prominence last year. DeepSeek became notable for breaking through globally in that period and is expected to release a new-generation model in the coming days, a development that the market is watching given the global tech share selloff the company triggered a year ago.
Alibaba’s messaging centers on improving capability per unit of inference cost while adding autonomy to application workflows. The company also referenced prior responses to competitive pressure: last year it released Qwen 2.5-Max, which it said was superior to a notable model from DeepSeek.
As Chinese AI vendors continue to iterate, Alibaba’s announcement highlights both the technical metrics it believes matter to enterprise customers - cost efficiency and large-workload throughput - and product-level features that support autonomous actions across apps. How users and enterprises translate those claims into sustained engagement remains to be seen amid active competition and impending model releases from rivals.