开源 OCR 成本 18 个月降 5 倍 百万页仅需 130 美元
Hugging Face 的 Daniel van Strien 在相同 GPU 环境下对比了 8 个开源 OCR 模型,发现处理 1000 页的成本从 2025 年 4 月 RolmOCR 的约 51 美分降至 2026 年 10 月 LightOnOCR-3-0.8B 的不足 10 美分,18 个月内成本下降约 5 倍。按此推算,数字化 100 万页的 GPU 租用成本已降至 65 至 130 美元区间,古籍与档案大规模文本化门槛显著降低。
报道正文
OCR models now digitize a million pages for under $130 as costs drop fivefold in 18 months
[Jonathan Kemper](https://the-decoder.com/author/jonathan-kemper/) [View the LinkedIn Profile of Jonathan Kemper](https://www.linkedin.com/in/jonathankemper/) Nov 19, 2026 Pulling text out of scanned books and records has gotten a lot cheaper with freely available AI models. Daniel van Strien 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 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.
According 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.

The 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.