The 🤗 Transformers library provides a centralized model‑definition framework that supports text, computer‑vision, audio, video, and multimodal models for both inference and training. It enables a single model definition to be compatible with major training stacks (Axolotl, Unsloth, DeepSpeed, FSDP, PyTorch‑Lightning) and inference engines (vLLM, SGLang, TGI, etc.) as well as auxiliary toolchains (llama.cpp, mlx). The project targets Python 3.10+ and PyTorch 2.6+, and can be installed via pip or uv, or cloned from GitHub for the latest development code. The high‑level Pipeline API abstracts preprocessing and post‑processing for tasks such as text generation (e.g., Qwen/Qwen2.5‑1.5B), conversational chat (Meta‑Llama‑3‑8B‑Instruct), automatic speech recognition (Whisper‑large‑v3), image classification (DINO‑v2‑small), and visual question answering (BLIP‑2). Users instantiate a pipeline, specify a model, and obtain cached results, reducing compute overhead. The library also hosts over 1 M pretrained checkpoints on the Hugging Face Hub, offers examples for each architecture, and exposes model internals consistently to lower entry barriers for researchers and developers. A citation‑ready paper (“Transformers: State‑of‑the‑Art Natural Language Processing”) documents the library’s design and impact.