Unlimited OCR: 3B open-source model reads 100-page PDFs in one pass, but scores just 46.2% on ParseBench
solyarisoftware · x · 2026-10-12
Jerry Liu (LlamaIndex) highlighted Unlimited OCR, an open-source 3B-parameter model that parses 100-page PDFs in a single 32K-context pass instead of page-by-page splitting, runs fully locally, and supports Transformers, vLLM, SGLang, Ollama and more.
But his ParseBench results temper the hype: it scores 46.2%, ranking #115. It's good at reading order and layout, slightly better than PaddleOCR on tables, but ignores visual elements, formatting and complex tables. PaddleOCR-VL-1.6 hits 67.4 and Qwen3.8 27B (Thinking) 70.8. Still a milestone given cloud OCR costs $1.50–15 per 1,000 pages, but the frontier has plenty of room to improve.
More from Models
- ChatGPT glitches on iPhone 15 Pro whenever it renders image comparisons — adam0672 · 2026-10-12
- Claude reportedly planned and coded a full video production in 2.5 hours — ZeroStateReflex · 2026-10-12
- KV caching explained: why it shapes GPU memory and drove the shift to GQA and MLA — blaizedsouza · 2026-10-12
- Claude Opus 5.5 vs GPT-6.1 Sol hands-on: this time GPT wins the vibe-check — CodeByPoonam · 2026-10-12
- DeepMind's AlphaProof Nexus reportedly cracks 9 Erdős problems, 44 OEIS conjectures — kimmonismus · 2026-10-12
- Codex keeps stalling on legit debugging over cybersecurity guardrails, user complains — Al_Grigor · 2026-10-12