Hugging Face has released the first RC version, v5.0.0rc0, of its renowned Transformers library, marking a significant leap from version 4 to 5 after five years of development. With over 3 million daily downloads and exceeding 1.2 billion installations, Transformers has cemented its status as the leading AI infrastructure library. The new version transitions to PyTorch as its sole backend, enhancing simplicity, modularity, and interoperability across various architectures—now more than 400. Key features include a focus on pre-training, improved training methods, and advanced inference capabilities with optimized kernels and APIs. The library simplifies model management through a strong modular approach, enhances quantization standards, and supports diverse inference engines. Hugging Face aims to remain at the forefront of AI innovation by continually adapting and expanding its open-source framework. Explore the release on GitHub here and learn more on the official Hugging Face blog.