Alibaba Cloud has open-sourced more than 100 of its newly launched AI models, collectively known as Qwen 2.5. The announcement was made during the company’s annual Apsara Conference.
The cloud computing arm of Alibaba Group has also unveiled a revamped full-stack infrastructure designed to meet the surging demand for robust AI computing. This new infrastructure encompasses innovative cloud products and services that enhance computing, networking, and data centre architecture, all aimed at supporting the development and wide-ranging applications of AI models.
Eddie Wu, Chairman and CEO of Alibaba Cloud Intelligence, said: “Alibaba Cloud is investing, with unprecedented intensity, in the research and development of AI technology and the building of its global infrastructure. We aim to establish an AI infrastructure of the future to serve our global customers and unlock their business potential.”
The newly-released Qwen 2.5 models range from 0.5 to 72 billion parameters in size and boast enhanced knowledge and stronger capabilities in maths and coding. Supporting over 29 languages, these models cater to a wide array of AI applications both at the edge and in the cloud across various sectors, from automotive and gaming to scientific research.
Alibaba Cloud’s open-source AI models gain traction
Since its debut in April 2023, the Qwen model series has garnered significant traction, surpassing 40 million downloads across platforms such as Hugging Face and ModelScope. These models have also inspired the creation of over 50,000 derivative models on Hugging Face alone.
Jingren Zhou, CTO of Alibaba Cloud Intelligence, commented: “This initiative is set to empower developers and corporations of all sizes, enhancing their ability to leverage AI technologies and further stimulating the growth of the open-source community.”
In addition to the open-source models, Alibaba Cloud announced an upgrade to its proprietary flagship model, Qwen-Max. The enhanced version reportedly demonstrates performance on par with other state-of-the-art models in areas such as language comprehension, reasoning, mathematics, and coding.
The company has also expanded its multimodal capabilities with a new text-to-video model as part of its Tongyi Wanxiang large model family. This model can generate high-quality videos in various visual styles, from realistic scenes to 3D animation, based on Chinese and English text instructions.
Furthermore, Alibaba Cloud introduced Qwen2-VL, an updated vision language model capable of comprehending videos lasting over 20 minutes and supporting video-based question-answering. The company also launched an AI Developer, a Qwen-powered AI assistant designed to support programmers in automating tasks such as requirement analysis, code programming, and bug identification and fixing.
To support these AI advancements, Alibaba Cloud has announced several infrastructure upgrades, including:
- CUBE DC 5.0, a next-generation data centre architecture that increases energy and operational efficiency.
- Alibaba Cloud Open Lake, a solution to maximise data utility for generative AI applications.
- PAI AI Scheduler, a proprietary cloud-native scheduling engine for enhanced computing resource management.
- DMS: OneMeta+OneOps, a platform for unified management of metadata across multiple cloud environments.
- 9th Generation Enterprise Elastic Compute Service (ECS) instance, offering improved performance for various applications.
These updates from Alibaba Cloud – including the release of over 100 open-source models – aim to provide comprehensive support for customers and partners to maximise the benefits of the latest technology in building more efficient, sustainable, and inclusive AI applications.
Source: AI NEWS
Recent Comments