unlimited-ocr
unlimited-ocr is Unlimited OCR Works: Welcome the Era of One-shot Long-horizon Parsing. - baidu/Unlimited-OCR. It is ranked #349 on the Dev Radar, in Other, first seen 48 days ago and shared in 3 posts (5k views).
GitHub - baidu/Unlimited-OCR: Unlimited OCR Works: Welcome the Era of One-shot Long-horizon Parsing. · GitHub Skip to content Navigation Menu Sign in Appearance settings Search / Sign in Sign up Appearance settings You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session. You switched accounts on another tab or window. Reload to refresh your session. Dismiss alert {{ message }} Repository navigation Unlimited-OCR Public Notifications You must be signed in to change notification settings Fork 3k ( 3k ) Star 28.3k ( 28.3k ) You must be signed in to star a…
What people said about unlimited-ocr on X
百度开源了一个 OCR 模型，光上个月在 HuggingFace 上就被下载了 165 万次，名字却没多少人听过。 它叫 Unlimited-OCR，3B 参数，搁自己电脑上就能跑。 传统 OCR 处理 PDF，往往是一页页切开识别，页与页之间的上下文很容易断掉。 Unlimited-OCR 换个打法：整份文档的页面一次性喂进去，连续做多页解析。 官方给这招起了个名，叫"一次性长视野解析"。 上下文窗口拉到了 32K。 PDF 先按 300dpi 转成图片，再整批送进模型。 单张图片有两种模式。 gundam 切块处理，base 按整页处理，多页和 PDF 只用 base。 部署门槛不算高。…
— @bkdgiffug, 1 days ago · 44 likes · see the post
长PDF要是拆成单页再往一块拼，跨页表格和阅读顺序十有八九会乱套，这事你八成也烦过。 百度开源的Unlimited-OCR打算多页一起塞进去：参数约30亿、激活量压得更低，论文里在32K上下文下能一口气解几十页，OmniDocBench上的成绩比DeepSeek-OCR那条基线还高。 它跑在本机，权重公开。Transformers、vLLM可以对接，社区那边也在跟进llama.cpp。云OCR按页数收钱，这东西等于把开销换成了机器和时间。 至于100页能不能一口吞完，得看显存和文档密度，仓库可没给你写保底。合同、论文先在本地过一轮没问题，终稿仍旧得人盯着。…
— @bkdgiffug, 5 days ago · 13 likes · see the post
百度发布并开源 Unlimited-OCR 2，模型约 30 亿参数，推理时仅激活约 5 亿参数。 6 月推出的第一版已经能连续解析约 40 页文档，但随着页数增加，每生成一个 token 仍要关注所有页面的视觉信息，导致解码越来越慢。 第二版引入 SR-SWA 稀疏注意力机制。模型不再每次都查看全部页面的视觉信息，而是重点关注当前页面，再挑选少量相关的视觉信息，结合最近生成的文字继续识别。百页测试中，单步解码比旧版 R-SWA 注意力方案快约 5.6 倍，比传统全注意力快约 7 倍。 显存节省也很明显。在百页文档测试中，通过分块解码和跨页缓存复用，单路推理的峰值 KV 缓存显存从 47GB 降至 1.77GB。另一组 16…
— @0xLogicrw, 9 days ago · 4 likes · see the post
Alternatives to unlimited-ocr
- kimi-k3.1-pro — Releasing the model weights of Kimi K3.1 Pro
- modelscope.ai — 📄 Paper
- deepseek-v4.1-flash-1017-gguf — Image-Text-to-Text model with 748B parameters. Download, explore files and run it on Hugging Face.
- deepseek-v4.1-flash-1017-uncensored-gsq-rco-gguf — Text Generation model with 552B parameters. Download, explore files and run it on Hugging Face.
- qwen3.8-flash-next-gguf — Image-Text-to-Text model with 177B parameters. Download, explore files and run it on Hugging Face.
- imagination-studio-by-keys-wasserman — @samael_voolt @SamJWasserman @ComfyUI @robinjhuang @hellorob
unlimited-ocr in numbers
- Rank on the Dev Radar: #349 of 9369
- Shared in 3 posts by 2 accounts: @bkdgiffug, @0xLogicrw
- 5k views on those posts
- First seen 48 days ago, last shared 1 days ago
- Pricing seen by AGK Intelligence: open source
- Market: Other
FAQ
What is unlimited-ocr?
Unlimited OCR Works: Welcome the Era of One-shot Long-horizon Parsing. - baidu/Unlimited-OCR It was first shared on X 48 days ago and is ranked #349 on the Dev Radar.
Is unlimited-ocr free?
It is open source.
Who shared unlimited-ocr?
2 accounts on X, including @bkdgiffug, @0xLogicrw, in 3 posts totalling 5k views.
Every post is read and classified by AGK Intelligence: what it is, which market it belongs to, and whether the link is a real tool. 31.7k posts from 5k X accounts over the last 21 days, 9.4k tools, 12 markets. New posts are collected several times an hour and the full ranking is recomputed every hour — last update 2026-11-05 14:03 UTC. Full method.