Why did we open-source our inference engine? Read the post

← Catalog

superlinked/docling-artifacts

Open comparison →

Primitive: /extract · Extract · Docling

This repository is an immutable composite artifact root for Docling's standard PDF pipeline. It contains only the assets exercised by Superlinked's current `docling` default and OCR runtime profiles:

MultimodalOCR-Document

Overview

Hardware: — drives latency, throughput & cost

Size80M params
Tasks /extract
Licenseother
Latency13.2 s
Throughput380 tok/s
Cost$0.584 /1M tok

Cost is approximate — computed from list GPU prices; your actual price depends on the provider you deploy SIE with.

Extraction

Output kindsParsed Document
Inputsimage · document
Max sequence length—

Benchmarks

olmOCR-Bench

general ocr en

Document text extraction accuracy across arxiv math, old scans, multi-column layouts, and tables

Corpus: 1,403 Queries: 1,403
ocr_bounded20
Performance L4 b1 c4
Extract 0.3 mpix/s
Extract p50 13.2s
ocr
Quality
accuracy 0.3347
default
Quality
accuracy 0.3170
Reference →

Open source inference for agents

Open-source inference for the models behind your agents. Run it yourself, or let us run it for you.

Contact us

Tell us about your use case and we'll get back to you shortly.

Apply for an inference grant

Free capacity on our hosted cluster for selected projects. Tell us what you run and we reply by email.