Alibaba Cloud, the digital technology arm of Alibaba Group, has unveiled an array of new AI models, tools, and infrastructure upgrades aimed at bolstering the global AI community. Announced at its annual developer summit, the enhancements include advanced large language models, innovative development tools, upgraded computing resources, and new support initiatives for developers and startups worldwide.
Empowering Developers with Advanced AI Models & Tools
Key to the announcement is the expansion of Alibaba Cloud’s proprietary large language model family, Qwen, with the release of the Qwen2.5 series, offering parameters ranging from 7 billion to 72 billion. These models are accessible via APIs on the Model Studio platform. Other multimodal AI models, such as Qwen-VL for vision understanding, Wanx2.1 (Tongyi Wanxiang) for visual generation, and Qwen-Audio for audio processing, are now available to developers globally.
Developers can also leverage Tongyi Lingma, an AI coding assistant powered by Qwen2.5-coder, to optimize coding efficiency with features like code completion, debugging, and batch unit test generation. Additional tools include Workflow, for breaking down complex tasks, and Agent, which supports multi-agent task collaboration. Enhanced tools such as Retrieval-Augmented Generation (RAG), Batch Reasoning, and AutoEval will be rolled out by the end of the month, further enhancing the reliability and efficiency of generative AI projects.
Infrastructure Upgrades for Seamless AI Development
To support intensive AI workloads, Alibaba Cloud introduced its 9th Generation Enterprise Elastic Compute Service (ECS) instance, slated for global release in April. This upgraded service boasts a 20% boost in computing efficiency and up to a 50% improvement in high-performance tasks through accelerated networks. The Alibaba Cloud Container Compute Service (ACS), designed for streamlined containerized workload deployment, will be available internationally starting in January 2025, reducing infrastructure complexity for developers.
Fostering Innovation with the GenAI Empowerment Program
Alibaba Cloud also launched the GenAI Empowerment Program, offering developers free cloud credits, training workshops, and marketing opportunities to accelerate the development of generative AI applications. The program aims to nurture innovation and connect participants with a broader ecosystem of AI pioneers.
Global Adoption of Qwen Models
Companies like Axcxept in Japan and OxValue.AI from the University of Oxford have already leveraged Alibaba Cloud’s Qwen models for transformative applications. Axcxept’s EZO model, based on Qwen2.5, excels in Japanese-specific AI tasks, while OxValue.AI uses Qwen-based multimodal models for precise company valuations.
These initiatives reaffirm Alibaba Cloud’s commitment to empowering developers and fostering innovation, paving the way for cost-effective and impactful AI-driven solutions worldwide.