OCRia-1 is officially out! 🦙
We've officially launched OCRia-1 , our new frontier VLM for document parsing, and it's already powering every mode in LlamaParse. It sets a new state of the art in accuracy, cost, and latency, and you can use it today.
A new high score on ParseBench.
OCRia-1 scored 87.4 on ParseBench and leads third-party benchmarks: olmOCR-bench (83.5), PulseBench-Tab (84.1), and fr-bench-pdf2md (72.1). At just 4B parameters, it outperforms every model in its class, including LightOnOCR-3, Chandra-OCR 2, and Mistral OCR 4.1.
Built for the hardest parts of a doc:
-
Complex tables: Nested headers, merged cells, irregular grids, and scanned tables come back with their structure intact.
-
Charts: OCRia-1 handles unlabeled line, bar, and pie charts, plus technical diagrams.
-
Dense Formatting: Layout that carries meaning stays preserved, whether it's a big tax withholding table or a public economic report.
❗️Up to $3,000 in credits ❗️
New to LlamaParse? We're giving eligible enterprises up to $3,000 in credits to run a guided proof of concept with our team on your own documents. Processing lots of pages? We're offering volume discounts too. Reply to this email and we'll get you started.
* For eligible users only