Unknown · depuis Nicht eindeutig belegt (Blog-Post „Today we're releasing“ ohne Datum im Suchergebnis; laut Suchmetadaten wenige Tage alt · 3× · vu le 15 oct. 2026
48
Momentum
Indra OCR is an OCR package from Datalab built around a 1-million-parameter recognition model. It converts PDFs and images into markdown, plain text or JSON with bounding boxes and runs on CPUs without a GPU or PyTorch. The code is licensed under Apache 2.0, the model weights under a modified OpenRAIL-M license.
Historique du momentum
18.07.16.10.
Fonctionnalités
| Key Benchmark (%) | OmniParseBench text: 41.3 (PaddleOCR PP-OCR 35.8; Tesseract 19.3); headline score 62.1 (PP-StructureV3 58.4) |
| License | Code: Apache 2.0; model weights: modified OpenRAIL-M (free for research, personal use and startups under $20M funding/revenue; no competing use with Datalab API) |
| Multimodality | No – pure OCR of PDFs/images with layout and line boxes (no multimodal LLM capability documented) |
| Platform | Python 3.10+, Linux (x86_64, aarch64) and macOS (Apple silicon); CPU, no GPU required; install via pip install indra-ocr |
| Release Date | Not clearly documented |