Datalab's Indra OCR: a 1M-parameter open model that hits 44 pages/s on CPUs, beating PaddleOCR
generativist · x · 2026-10-22
Datalab open-sourced Indra OCR on Hugging Face: a 1M-parameter model running on CPU via ONNX Runtime (no PyTorch/GPU). It scores 66.7 on OmniParseBench (71.4 with --math) vs PaddleOCR PP-StructureV3's 58.4, and its text score 46.7 beats PP-OCR (35.8) and Tesseract (19.3). It runs 44 pages/s on 32 cores — 15x Tesseract, 10x PP-OCR — supports 80+ languages, handwriting, scans tilted up to 45°, layout-aware markdown/JSON output, and OCRs only what's missing in digital PDFs.
More from Models
- Recursive AI ships Flint, a 40B-A2.5B MoE claimed to match Qwen3.6-27B, weights on Hugging Face — ChengleiSi · 2026-10-22
- DeepSeek ships V4.1-Flash-1017, now the #1 open model on RSI-Inference Bench — scaling01 · 2026-10-22
- OpenAI's Cybersecurity Guardrails Block a Dev From Auditing His Own Patched Server — Next_Technology6361 · 2026-10-22
- TPN-006 hits Hugging Face: 7.7GB compressed model beats its 17GB Mistral teacher — markjeffrey · 2026-10-22
- Kimi K3.1's latest checkpoint draws 49k-token SVG portrait zero-shot in 22 minutes — lyraxana · 2026-10-22
- MiniMax launches M3.1 Flash at 1/3 the cost and 3x the speed of M3, weights out Oct 23 — MiniMax_AI · 2026-10-22