📄 中文摘要
2025年，亚马逊SageMaker AI在核心基础设施方面取得了显著进展，主要体现在四个方面：容量、价格性能、可观察性和可用性。推出的灵活训练计划显著提升了容量，使得用户能够根据需求调整训练资源。此外，推理工作负载的价格性能也得到了改善，降低了成本，提高了效率。这些改进为用户提供了更大的灵活性和更高的性价比，推动了AI应用的进一步发展。后续部分将探讨可观察性、模型定制和模型托管等方面的增强。
2025年亚马逊SageMaker AI回顾第一部分：灵活的训练计划与推理工作负载的价格性能改进
📄 English Summary
Amazon SageMaker AI in 2025, a year in review part 1: Flexible Training Plans and improvements to price performance for inference workloads
In 2025, Amazon SageMaker AI achieved significant advancements in core infrastructure across four key dimensions: capacity, price performance, observability, and usability. The introduction of Flexible Training Plans greatly enhanced capacity, allowing users to adjust training resources based on their needs. Additionally, improvements in price performance for inference workloads reduced costs and increased efficiency. These enhancements provided users with greater flexibility and better value, driving further development in AI applications. Subsequent parts will discuss enhancements in observability, model customization, and model hosting.