Alibaba Cloud, the cloud computing arm of Alibaba Group, presented its next wave of artificial intelligence (AI) innovations at the Apsara Conference 2025. The company highlighted new large language models, multimodal systems, development platforms, and infrastructure designed to help enterprises and developers build AI-powered applications at scale.
“In the future, large AI models will be deeply integrated into a wide range of devices, functioning like operating systems — equipped with persistent memory, seamless cloud-edge coordination, and the ability to continuously evolve,” said Eddie Wu, chair and CEO of Alibaba Cloud Intelligence.
Wu added that Alibaba will continue to open-source its Qwen models to empower global developers.
“We remain committed to shaping Qwen into the operating system of the AI era,” Wu said, noting the company’s RMB 380 billion investment plan in AI and cloud infrastructure over the next three years.
Alibaba’s Qwen series, first introduced in 2023, has since grown into one of the most widely used open-source AI families. More than 300 models have been released under Qwen and Wan, downloaded over 600 million times. Over one million users have accessed Qwen on Model Studio, Alibaba’s AI development platform.
At the conference, Alibaba unveiled Qwen3-Max, its most advanced large language model with more than one trillion parameters. It is designed for both instruction-based use and reasoning tasks, with applications in code generation and intelligent agents. Benchmarks show strong performance in software troubleshooting and conversational agent capabilities. Complementing it are Qwen3-VL, a vision-language model for flexible deployment, and Qwen3-Omni, a multilingual system that processes text, images, audio, and video with real-time responses.
Specialized models such as Qwen3-Coder, Qwen3-Image-Edit, and the Fun speech series were also updated, while Wan2.5 previewed new video and image generation features. Together, these models expand Alibaba’s AI portfolio for both enterprise and consumer applications.
To help organizations deploy AI agents, Alibaba expanded Model Studio with an Agent Development Kit and enterprise-grade features such as data fusion, dynamic scheduling, and sandboxing. More than 800,000 agents have been created through Model Studio for uses ranging from content creation and marketing to smart home management.
The company also enhanced AgentBay, a cloud-based platform for multimodal agents, and launched Lingyang AgentOne, a one-stop AI application suite for enterprises. For consumers, its Quark division introduced Zaodian, a platform for AI-driven video and image creation powered by Wan models.
Beyond applications, Alibaba emphasized the role of infrastructure in enabling AI at scale. The company rolled out upgrades in storage, networking, and databases, along with new security frameworks and container technologies. It also enhanced its Platform for AI (PAI), supporting large-scale training and deployment of advanced models.