Hi @mkesper! I'm Dosu and I’m helping the docling team.

There are no official minimum or recommended hardware requirements documented for Docling; requirements depend heavily on your workload, models, and document sizes. However, based on user reports and issue discussions, here are practical estimates:

For CPU mode, typical memory usage ranges from 2–4 GB RAM for standard pipelines and OCR tasks, but can spike much higher—up to 80 GB RAM—when processing large documents or using complex models. For GPU mode, you’ll need an NVIDIA GPU with CUDA support. VRAM usage varies: standard pipelines may use 1–4 GB VRAM, but advanced models (like code/formula or vision models) can require 6–21 GB VRAM. Some users report VRAM usage exceeding 10 GB and persistent memory growth when processing multiple files, so a GPU with at least 12–24 GB VRAM is recommended for demanding workloads. Example breakdowns for specific components are available here.

Memory usage can increase over time due to leaks or caching, so monitoring and chunking large files is advised (example). For Docker deployments, ensure your VM has enough RAM allocated to avoid out-of-memory errors (example).

Ultimately, requirements will vary—monitor your resource usage and adjust accordingly, especially for large-scale or high-throughput scenarios.

To reply, just mention @dosu.

Docs are dead. Just use Dosu.