Baidu open-sources Unlimited-OCR: 3B model reads 100-page PDFs in one shot
udmrzn · x · 2026-11-23
Baidu released Unlimited-OCR, a 3B-parameter open-source (MIT) OCR model that parses entire 100-page PDFs in a single pass with a 32K context window, instead of chopping documents page by page.
Key facts:
- Scores 93% on OmniDocBench v1.5 (+6 over the DeepSeek-OCR baseline); edit distance stays under 0.11 past 40 pages
- Multilingual, runs 100% locally, works with Transformers, vLLM, SGLang, and Docker
- Versus cloud OCR (Textract, Google Vision, Azure) at $1.50–15 per 1,000 pages, local means free
- 32k GitHub stars, millions of HF downloads; Unlimited OCR 2 is live on Baidu Cloud (TextMind) with a paper on arXiv
Related event: Baidu Open-Sources Unlimited OCR2, Reads Hundreds of Pages at Once(3 posts)→
More from Models
- Pedro Domingos: OpenAI admits pretraining has barely improved since 2022 — pmddomingos · 2026-11-23
- Inkling-2-Mini hits 10.04x on KernelBench-Mega locally, beating Grok 4.7 and GPT-6.1 — ziqiao_ma · 2026-11-23
- Vocab Surgery: editing 10 dictionary words drops a model's refusal rate to near zero — morgymcg · 2026-11-23
- Perplexity across tokenizers: researcher pushes back on Anthropic comparison controversy — quantum_geoff · 2026-11-23
- Thinking Machines releases Inkling 2 Small, scoring 31 on AA Intelligence Index — _Hao_Zhu · 2026-11-23
- Leaked OpenAI loss ladder: unknown capabilities await at 0.68-0.7 — tallmetommy · 2026-11-23