Today we’re releasing Indra OCR, an OCR package built around a 1M-parameter recognition model. It reads 49 pages per second on 32 CPU cores, 11-17x faster than Tesseract and PaddleOCR’s PP-OCR text models. Single page latency is ~230 ms. And it beats both of them on OmniParseBench text benchmark. It isn’t only text. Indra OCR gives you the full layout: a bounding box for every line and every layout block, block labels (headings, lists, tables, captions and more) and reading order, as markdown or...