LlamaIndex releases OCRia-1 document parsing model, powering LlamaParse modes
Article timeline
The articles in this story. Times are ET.
Jerry LiuXLlamaIndex positions itself as an applied research lab for document OCR and agentsAILlamaIndex says it is an applied research lab building models and agents for document work, with OCR models available now in LlamaParse. It cites its OCRia-1 4B VLM for document parsing and Extract v2.6 launched in the last two weeks. The company says ParseBench has 150+ models and 30k+ monthly downloads.
- Jerry LiuXLlamaIndex releases OCRia-1, a 4B document parsing model for complex tables
AILlamaIndex says its OCRia-1, a 4B-parameter document parsing model, scores 91.5% on ParseBench, 92.9% on olmOCR-bench, 84.1% on PulseBench-Tab, and 83.2% on fr-bench-pdf2md for table accuracy. The model handles nested headers, merged cells, and borderless tables, and it powers LlamaParse's cost-effective, agentic, and agentic plus modes.
- Jerry LiuXLlamaIndex releases OCRia-1, a 4B document parsing model with a 37-page report
AILlamaIndex releases OCRia-1, a 4B-parameter document parsing model, and publishes a 37-page technical report on it. The company says it leads ParseBench with a score of 87.4 and decodes 2x to 4x faster than recent open-weight models. The report compares OCRia-1 against Infinity-Parser, LightOnOCR 3, Chandra 2, and dots.mocr on olmOCR-bench, PulseBench-Tab, and fr-bench-pdf2md.
- Jerry LiuXLlamaIndex releases OCRia-1, a 4B document parsing model
AILlamaIndex releases OCRia-1, a 4B-parameter vision-language model for document parsing, and says it scores 87.4 on ParseBench. The company reports leading scores on olmOCR-bench (83.5 overall, 92.9 on tables), PulseBench-Tab (84.1), and fr-bench-pdf2md (72.1), and says it outperforms similarly sized models such as LightOnOCR-3 and Chandra-OCR 2. OCRia-1 now powers all LlamaParse modes, and enterprises that book a demo can get up to $3k in credits for evaluation.
- Jerry LiuXOCRia-1 document parsing model now available on OpenDocRouter
AILlamaIndex has added its OCRia-1 document parsing model to OpenDocRouter, where users can compare it against other parsers on accuracy, speed, and cost. Requests set the model to llamaindex/ocria-1, and new accounts start with $5 in credit. OCRia-1 is also available through LlamaParse.
LlamaIndex 🦙OfficialLlamaIndex launches OCRia-1 document parsing model on OpenDocRouterAILlamaIndex launched OCRia-1, a document parsing model that powers LlamaParse, and made it available on OpenDocRouter. Users can run it against other parsing models on their own documents to compare accuracy, speed, and cost, by setting the model to llamaindex/ocria-1 in a parse request. New OpenDocRouter accounts receive $5 in credit.
- LlamaIndex 🦙OfficialLlamaIndex releases OCRia-1, a 4B document parsing model with top benchmark scores
AILlamaIndex introduces OCRia-1, a 4B-parameter vision-language model for document parsing that scores 87.4 on ParseBench and leads olmOCR-bench (83.5 overall, 92.9 on tables), PulseBench-Tab (84.1), and fr-bench-pdf2md (72.1). The company says the model handles nested headers, merged cells, scanned tables, and unlabeled charts, and it now powers LlamaParse's cost-effective, agentic, and agentic plus modes. LlamaIndex is offering volume discounts and up to $3,000 in credits to enterprises that book a demo, with a promotion running for 48 hours.
- Jerry LiuXLlamaIndex releases OCRia-1, a 4B document-parsing VLM that tops ParseBench
AILlamaIndex releases OCRia-1, a 4B-parameter vision-language model for document parsing, which scores 87.4 on ParseBench. It leads three third-party benchmarks: 83.5 overall on olmOCR-bench, 84.1 on PulseBench-Tab, and 72.1 on fr-bench-pdf2md. The model powers all LlamaParse modes, and enterprises booking a demo can receive up to $3k in credits.
- Jerry LiuXLlamaIndex releases OCRia-1, a 4B document parsing model with SOTA scores
AILlamaIndex releases OCRia-1, a 4B-parameter vision-language model for document parsing, reporting 87.4 on ParseBench. It also scores 83.5 overall and 92.9 on tables in olmOCR-bench, 84.1 on PulseBench-Tab, and 72.1 on fr-bench-pdf2md, and the company says it outperforms similar-sized models including LightOnOCR-3, Chandra-OCR 2, and Mistral OCR 4.1. OCRia-1 now powers LlamaParse's cost-effective, agentic, and agentic plus modes, and enterprises that book a demo can get up to $3k in credits for evaluation.
