heron is the default layout analysis model of the Docling project, designed for robust and high-quality document layout understanding.
Runs On
What it takes to serve docling-layout-heron (43M parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.
| Precision | Weights | Memory needed | Cheapest setup | Per hour | Also fits |
|---|---|---|---|---|---|
| 16-bit | 0.1 GB | 0.1 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 8-bit | 0.0 GB | 0.1 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 4-bit | 0.0 GB | 0.0 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Sep 18, 2026.
SAVRN's Notes on docling-layout-heron
Before a language model reads a scanned permit or lease, something has to find the blocks on the page. That is heron's job as the Docling project's default layout analysis model, a 43M-parameter RTDetrV2ForObjectDetection network. At 16-bit the weights are 0.1 GB and 0.1 GB of memory is needed. The cheapest setup the SAVRN Index prices is one MI300X with 192 GB at $1.85 an hour on-demand, so it runs as a sidecar on the card already doing the reading, and we would not provision for it separately.
Apache 2.0 allows commercial use and redistribution with notices kept and changes stated, so a document pipeline built on it can be sold. Check the evidence: no evaluations are reported here, so the technical report at arXiv:2509.11720 and the Docling report at arXiv:2408.09869 are the only account of how it was measured. Released April 15, 2025, updated February 9, 2026.
Model Card
By Docling, published under apache-2.0, revision 8f39ad3c0b4c.
Document Layout Analysis "heron"
heron is the default layout analysis model of the Docling project, designed for robust and high-quality document layout understanding.
For an in-depth description of the model architecture, training datasets, and evaluation methodology, please refer to our technical report:"Advanced Layout Analysis Models for Docling", Nikolaos Livathinos et al., https://arxiv.org/abs/2509.11720
Inference code example
Prerequisites:
pip install transformers Pillow torch requests
Prediction:
Configuration
- Architecture
- RTDetrV2ForObjectDetection
- Stored precision
- float32
- Model type
- rt_detr_v2
Identity and Version
- Repository
- docling-project/docling-layout-heron
- Publisher
- Docling
- Task
- Not stated by the source
- Modality
- Other
- Library
- Not stated by the source
- Parameters
- 43M parameters
- Languages
- Not stated by the source
- Revision
- 8f39ad3c0b4c58e9c2d2c84a38465abf757272d8
- First published
- 2025-04-15
- Last updated
- 2026-02-09
Files and Weights
6 files, 171.8 MB in total. The weights are 1 file totalling 171.7 MB in safetensors.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| model.safetensors | Weights | 171.7 MB | 00333a434519 |
| config.json | Configuration | 3.3 KB | — |
| preprocessor_config.json | Configuration | 444 B | — |
| README.md | Documentation | 3.2 KB | — |
| docling_heron_400.png | Other | 96.9 KB | — |
| .gitattributes | Repository | 1.5 KB | — |
License and Download
- License
- apache-2.0
- Access
- Open weights, no gate
- Download size
- 171.7 MB
Released by Docling through its official repository on Hugging Face. Read the license.
Built From
- Described by arXiv:2408.09869
- Described by arXiv:2509.11720
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 171.7 MB |
| 16-bit | 0.1 GB |
| 8-bit | 0.0 GB |
| 4-bit | 0.0 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About docling-layout-heron
How much GPU memory does docling-layout-heron need?
About 0.1 GB at 16-bit and 0 GB at 4-bit: the weights (43M parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run docling-layout-heron on?
At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.
Can I use docling-layout-heron commercially?
Yes. docling-layout-heron is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.