Alibaba has announced a break through AI model, Qwen 3.5, which greatly enhances China’s competition with American companies who are involved in accelerating AI technologies. Qwen 3.5 includes the first iteration of the models in the Qwen 3 series released to the public as open-source, and is expected to produce “outstanding results across the board” when bench marking by industries. Moreover, Qwen 3.5 out performed several well-known Western models in various key benchmarks, including Open AI’s GPT-5, Anthropic’s Claude Opus 4, and Google’s Gemini 3 Pro.
The technical architecture for Qwen 3.5 has 397 billion total parameters; only 17 billion parameters are being utilized on each forward pass, which improves computational efficiency while providing high performance. Alibaba indicated that using this architecture optimizes processing speed while not limiting capabilities, providing developers with a more economically viable option for large-scale AI implementations.
Qwen 3.5 is 60% less expensive to operate, and eight times more resource effective when utilizing large workloads than previous iterations of the series. Qwen 3.5 also introduces visual-agentic functionality that enables Qwen 3.5 to perform tasks autonomously within mobile applications as well as on desktop applications — creating a heightened focus on the development of agent-based AI systems worldwide.
According to a statement from Alibaba that was reported on by Reuters, Qwen 3.0 has been developed to provide developers and businesses with the ability to increase the speed at which they are able to deliver new software and service applications. Additionally, the product will enable them to perform more tasks using the same amount of computing power, therefore establishing a new standard in terms of performance per dollar spent on machine learning.
Recent announcements of new artificial intelligence products from companies such as Doubao’s and Seedances’ improved versions of their chatbots and video creation tools, respectively, have shown that there is a great deal of activity taking place within the marketplace. Furthermore, other companies, including Zhipu’s, MiniMax’s, and Moonshot’s, have launched newer and more sophisticated models. As well, DeepSeek has begun accepting pre-orders for its highly awaited version (V4) model, which should be available within the next month; therefore, these two developments alone will make the global artificial intelligence landscape extremely competitive.