256M VLM for end-to-end document AI
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
SmolDocling integrates with Hugging Face's comprehensive ML stack (Transformers, Diffusers, PEFT) allowing developers to combine document parsing with fine-tuning, inference optimization, and deployment on Inference Endpoints or Spaces.
SmolDocling can be hosted and versioned on the Hugging Face Hub, enabling teams to collaborate on model improvements, share datasets, and manage access controls through Team or Enterprise plans with SSO, audit logs, and resource groups.
SmolDocling, as part of the Hugging Face ecosystem, enables developers to parse and extract document content using state-of-the-art models. Users can leverage the Hub's 2M+ models and integrate document parsing into ML pipelines via Python libraries like Transformers and the Hub client library.
SmolDocling models can be deployed via Hugging Face Inference Endpoints (starting at $0.60/hour for GPU) or integrated into Spaces applications, providing managed, scalable document processing without infrastructure management.