Docling
Docling is IBM’s MIT document parser: pip install docling, point it at a PDF or Office file, and get structured DoclingDocument (markdown, JSON, tables,…
OUR TAKE
Pick Docling for engineers who will run MIT Docling on a laptop or cluster. You bring hardware.
Good fit
Engineers who will run MIT Docling on a laptop or cluster.
Poor fit
Teams that need a hosted $/page SLA (AWS Textract) without operating Python.
Self-host
USD, before tax · Checked 2026-10-28 · github.com/docling-project/docling
OUR TAKE
Pick Docling for engineers who will run MIT Docling on a laptop or cluster. You bring hardware. A 200-page scan is not “free compute.”.
Good fit
Engineers who will run MIT Docling on a laptop or cluster.
Poor fit
Teams that need a hosted $/page SLA (AWS Textract) without operating Python.
Self-host
USD, before tax · Checked 2026-10-28 · github.com/docling-project/docling
What it costs
USD, before tax · Checked 2026-10-28 · github.com/docling-project/docling
Self-host
MIT license. You pay CPU/GPU
$0
Managed watsonx / IBM SaaS
Not published by the vendor
Not published by the vendor
Checked 2026-10-28·github.com/docling-project/docling
Before you pay
Good fit
Engineers who will run MIT Docling on a laptop or cluster.
Poor fit
Teams that need a hosted $/page SLA (AWS Textract) without operating Python.
Watch out
You bring hardware. A 200-page scan is not “free compute.”
Instead, try
MinerU if scientific PDF → markdown you can also host; Unstructured if Apache ETL with a $0.015 hosted page
Compare Docling with
Limits, in full
- 01
You bring hardware. A 200-page scan is not “free compute.”
- 02
No public hosted dollar. on the pages we opened.
- 03
A parser, not an invoice workflow. For SenseML insurance JSON see Sensible.
If Docling isn't it
Why: scientific PDF → markdown you can also host.
Free· paid from$9.90/mo
Free has no APIChecked 2026-10-28·mineru.io
vs Docling
Why: Apache ETL with a $0.015 hosted page.
Free· paid from$0.015
$0.015/page after 10kChecked 2026-10-22·unstructured.io/pricing
vs Docling
Why: Apache OCR/VL when you want detection boxes, not DoclingDocument.
Free· open source,Apache-2.0
The Space can be slowChecked 2026-10-28·github.com/PaddlePaddle/PaddleOCR
vs DoclingLimits & alternatives3 limits · MinerU, Unstructured, PaddleOCR instead
Limits, in full
- 01
You bring hardware. A 200-page scan is not “free compute.”
- 02
No public hosted dollar. on the pages we opened.
- 03
A parser, not an invoice workflow. For SenseML insurance JSON see Sensible.
If Docling isn't it
Why: scientific PDF → markdown you can also host.
Free· paid from$9.90/mo
Free has no APIChecked 2026-10-28·mineru.io
vs DoclingWhy: Apache ETL with a $0.015 hosted page.
Free· paid from$0.015
$0.015/page after 10kChecked 2026-10-22·unstructured.io/pricing
vs DoclingWhy: Apache OCR/VL when you want detection boxes, not DoclingDocument.
Free· open source,Apache-2.0
The Space can be slowChecked 2026-10-28·github.com/PaddlePaddle/PaddleOCR
vs Docling