At the Apsara Conference, Alibaba Cloud, the digital technology backbone of Alibaba Group, announced the release of over 100 large language models from its newly launched Qwen 2.5 series to the global open-source community, with a deadline set for full deployment by December 2024.
In conjunction with this release, Alibaba Cloud unveiled a revamped full-stack infrastructure designed to meet the growing demands for AI computing. The upgraded infrastructure includes cloud products and services to enhance computing, networking, and data center architecture, supporting the expansive development of AI models.
“Our goal is to create the AI infrastructure of the future to serve global customers and unlock their business potential,” said Eddie Wu, Chairman and CEO of Alibaba Cloud Intelligence.
Qwen 2.5: Expanding AI Capabilities
The newly-released Qwen 2.5 models, ranging from 0.5 to 72 billion parameters, offer enhanced capabilities in math and coding and support over 29 languages. The models cater to diverse applications across sectors like automobiles, gaming, and scientific research.
Since its debut in April 2023, the Qwen series has seen over 40 million downloads on platforms like Hugging Face and ModelScope. With the Qwen 2.5 release, Alibaba Cloud will make over 100 models open-source, including base, instruct, and quantized models, spanning multiple modalities such as language, audio, and vision.
New AI Tools and Multimodal Advancements
As part of its innovation in AI, Alibaba Cloud introduced an upgraded version of its proprietary Qwen-Max model, designed to excel in language comprehension, reasoning, math, and coding. The company also unveiled a new text-to-video model within the Tongyi Wanxiang family, capable of generating high-quality videos in diverse visual styles based on text input in both English and Chinese.
Additionally, the Qwen2-VL vision language model has been enhanced to process videos of up to 20 minutes, supporting video-based question-answering and sophisticated decision-making. This model is geared toward integration into mobile phones, automobiles, and robots for automated operations.
Strengthened AI Infrastructure for Global Needs
To support these AI advancements, Alibaba Cloud announced its next-generation data center architecture, CUBE DC 5.0, which improves energy efficiency and speeds up deployment times. The company also introduced the Open Lake solution for maximizing data utility in generative AI applications and launched the PAI AI Scheduler for integrated model training and inference.
These updates, along with the introduction of the 9th Generation Elastic Compute Service (ECS), aim to support customers and partners in building more efficient, sustainable AI applications by the end of 2024.
![[Latest] Global Data Center CPU Market Size/Share Worth USD 48.9 Billion by 2033 at a 15.2% CAGR: Custom Market Insights (Analysis, Outlook, Leaders, Report, Trends, Forecast, Segmentation, Growth, Growth Rate, Value)](https://energy.asia/wp-content/uploads/2024/12/global-data-center-cpu-market-2024-2033-by-billion--120x86.png)
