{"slug": "ocr-models-now-digitize-a-million-pages-for-under-130-as-costs-drop-fivefold-in", "title": "OCR models now digitize a million pages for under $130 as costs drop fivefold in 18 months", "summary": "OCR costs for digitizing scanned documents fell roughly fivefold in 18 months, with a million pages now costing between $65 and $130 in GPU rental, according to a benchmark by Daniel van Strien of Hugging Face. Van Strien ran eight open OCR models on 850 pages from old natural history books, the 1771 Encyclopaedia Britannica, and archival documents on a rented Nvidia RTX PRO 6000 at $2.75 per hour, finding the April 2025 RolmOCR model cost about 51 cents per 1,000 pages while the October 2026 LightOnOCR-3-0.8B came in just under 10 cents. Van Strien reported that optimizing how the models ran saved about as much money as switching to newer models, though some models hallucinated text on blank pages or repeated content.", "body_md": "# OCR models now digitize a million pages for under $130 as costs drop fivefold in 18 months\n\n**Pulling text out of scanned books and records has gotten a lot cheaper with freely available AI models.** [Daniel van Strien](https://danielvanstrien.xyz/posts/2026/ocr-cost-trend/) of Hugging Face compared eight such models that recognize text in images. He tested them on 850 pages from old natural history books, the 1771 Encyclopaedia Britannica, and typed and printed archival documents. All models ran on the same GPU, an [Nvidia RTX PRO 6000](https://the-decoder.com/china-orders-bytedance-and-alibaba-to-cancel-nvidia-ai-chip-purchases/) that van Strien rented through Hugging Face for $2.75 per hour. At the same image size, the oldest model from April 2025 (RolmOCR) cost about 51 cents per 1,000 pages. The newest from October 2026 (LightOnOCR-3-0.8B) came in at just under 10 cents. A million pages now runs between $65 and $130 in GPU rental, depending on the source material.\n\nAccording to van Strien, optimizing how the models ran saved about as much money as switching to newer models did, if not more. That meant feeding pages more evenly, using smaller scans, updating software, and running a faster GPU. The newest model read book pages at least as accurately as the oldest one. Some models, though, hallucinated text on blank pages or repeated the same content over and over.\n\nThe cost of scanning 1,000 pages has dropped from 51 cents for the oldest model to just under 10 cents for the newest one; the five models in between are almost on par with each other.\n\n```\nAI News Without the Hype – Curated by Humans\n\n\t\t\t\t\tSubscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive \"AI Radar\" frontier report six times a year, full archive access, and access to our comment section.\t\t\t\t\n\n\t\t\t\t\tSubscribe now\n```\n\n[Daniel van Strien](https://danielvanstrien.xyz/posts/2026/ocr-cost-trend/)", "url": "https://wpnews.pro/news/ocr-models-now-digitize-a-million-pages-for-under-130-as-costs-drop-fivefold-in", "canonical_source": "https://the-decoder.com/ocr-models-now-digitize-a-million-pages-for-under-130-as-costs-drop-fivefold-in-18-months/", "published_at": "2026-11-19 09:28:01+00:00", "updated_at": "2026-11-19 09:42:29.247900+00:00", "lang": "en", "topics": ["computer-vision", "ai-research", "ai-tools", "ai-infrastructure"], "entities": ["Daniel van Strien", "Hugging Face", "Nvidia RTX PRO 6000", "RolmOCR", "LightOnOCR-3-0.8B", "Encyclopaedia Britannica"], "also_reported_by": [], "takeaway": "Running eight open OCR models on a rented Nvidia RTX PRO 6000 at $2.75/hour, van Strien measured per-1,000-page costs falling from about 51 cents (RolmOCR, April 2025) to just under 10 cents (LightOnOCR-3-0.8B, October 2026), with runtime optimization saving as much as newer models.", "recommendation": {"advice": "Before switching OCR models for cost, optimize the serving setup first — even page feeding, smaller scans, updated software and a faster GPU saved as much as a newer model in van Strien's tests.", "playbook": "inference-serving", "playbook_url": "https://wpnews.pro/playbooks/inference-serving"}, "alternates": {"html": "https://wpnews.pro/news/ocr-models-now-digitize-a-million-pages-for-under-130-as-costs-drop-fivefold-in", "markdown": "https://wpnews.pro/news/ocr-models-now-digitize-a-million-pages-for-under-130-as-costs-drop-fivefold-in.md", "text": "https://wpnews.pro/news/ocr-models-now-digitize-a-million-pages-for-under-130-as-costs-drop-fivefold-in.txt", "jsonld": "https://wpnews.pro/news/ocr-models-now-digitize-a-million-pages-for-under-130-as-costs-drop-fivefold-in.jsonld"}}