PaddleOCR 3.5 Integrates Transformers Backend for OCR and Document Parsing

PaddleOCR 3.5 now supports Transformers as an inference backend for its OCR and document parsing models, including PP-OCRv5 and PaddleOCR-VL 1.5. This release introduces a flexible inference-engine interface, allowing developers to select the backend via the engine parameter and configure options like dtype, device placement, and attention implementation through engine_config. The integration aims to reduce friction for developers building RAG, Document AI, search, analytics, or agent applications within Hugging Face-centered stacks. While the Transformers backend offers a familiar development experience and Hub-compatible model discovery, the default paddle_static backend remains recommended for maximum throughput. A live demo is available on Hugging Face Spaces. The release was supported by Hugging Face engineers, including Anton Vlasjuk, Raushan Turganbay, and Yoni Gozlan.