Alibaba Cloud has revealed new AI models, tools, and infrastructure upgrades aimed at its international customers.
The enhancements are accessible via the company's availability zones in Singapore and feature the latest iterations of the proprietary large language model (LLM) series, known as Qwen.
These include the large-scale Mixture of Experts (MoE) model Qwen-Max, the reasoning model QwQ-Plus, the visual reasoning model QVQ-Max, and the end-to-end multimodal model Qwen2.5-Omni-7b.
The updated version of QwQ-Plus focuses on advanced reasoning, particularly analytical challenges such as complex QA tasks and math problems, providing precise algorithm-driven solutions. QVQ-Max is tailored for tackling complex multimodal issues, allowing for the integration of visual input with structured output, thus enhancing the reasoning capabilities of the model.
Selina Yuan, President of International Business at Alibaba Cloud Intelligence, commented on these developments: "We are launching a series of Platform-as-a-Service (PaaS) and AI capability updates to meet the growing demand for digital transformation from across the globe. These upgrades allow us to deliver even more secure and high-performance services that empower businesses to scale and innovate in an AI-driven world. As cloud and AI become essential for global growth, we are committed to enhancing our core product offerings to address our customers' evolving needs."
Alibaba Cloud's Platform for AI (PAI) includes significant improvements, with PAI-Elastic Algorithm Service (EAS) offering enhanced distributed inference capabilities through a multi-node architecture.
This change is designed to address the demands stemming from the growth of MoE structures and extensive text processing. Additionally, the introduction of the prefill-decode disaggregation function aims to boost performance, exemplified by a 92% increase in concurrency and a 91% increase in tokens per second (TPS) when used with the Qwen2.5-72B model.
Further updates to the PAI-Model Gallery provide users with access to nearly 300 open-source models, alongside Alibaba Cloud's proprietary models. Users benefit from a no-code deployment process and additional features such as model evaluation and distillation aimed at cost-effective deployment.
In terms of data management, Alibaba Cloud has incorporated its AI inference capabilities into its PolarDB database, designed to enhance efficiency by removing the need for data movement during inference workflows.
This integration aims to improve latency and data security, optimising performance for text-centric tasks and smart retrieval applications.
The AnalyticDB data warehouse has now been integrated into Model Studio, which is a platform for AI application development. This integration assists organisations in seamlessly connecting their knowledge bases to AI models and tools, simplifying the development of context-aware applications.
Additionally, several Software-as-a-Service (SaaS) AI products were introduced to further support industry transformations. These tools leverage AI to improve data analytics, enhance automation, and facilitate content creation, helping enterprises in fields such as e-commerce, gaming, and education.
The new AI Doc tool harnesses LLMs to efficiently process and analyse various document types, while the Smart Studio platform offers AI-powered content creation capabilities, supporting diverse use cases from marketing to personalisation of retail experiences.
Furthermore, Alibaba Cloud launched a new AI search function to aid enterprise clients, specifically SMEs, in identifying suitable cloud solutions and providing access to training resources.
The company has also revised its partner incentive policies to offer greater flexibility, higher commission rates, and better support for resellers and distributors, coupled with enhanced training and assistance, to strengthen their capabilities within the evolving digital ecosystem.
These developments align with Alibaba's previously announced investment of USD $53 billion over the next three years to advance its cloud and AI infrastructure, endorsed by its growing global presence spanning 87 availability zones across 29 regions.
