Run Baidu Unlimited-OCR (3B) on RTX 4060 8GB via SGLang & Transformers. Phone Camera OCR server with CPU offload
百度 Unlimited-OCR 模型的本地化服务封装：OpenAI 兼容 API + 单图/批量 OCR 端点
Unlimited-OCR 2026 – One-shot long-horizon document OCR from Baidu. 3B MoE (570M active), 93.92% on OmniDocBench. Read 40+ pages in one pass. Setup.exe included.
Convert PDF to markdown using a GUI/CLI for Unlimited-OCR. Uses GGUF models
Reference Sliding Window Attention (R-SWA) from Baidu's Unlimited OCR, reproduced from scratch in PyTorch. Runs on Apple Silicon (MPS) and CPU — no CUDA. Constant KV cache, with benchmarks.
Unlimited-OCR 的 OpenVINO 本地适配：显式 KV、R-SWA cache、12 层 sparse MoE decode、FP4/INT8 mixed experts
Unlimited-OCR: Free 3B local VLM for multi-page document parsing. 32K context window, 93% accuracy, single-pass processing via Ollama, vLLM & llama.cpp.
A high-performance document parsing and structuring pipeline using Baidu Unlimited-OCR and local Ollama.
Add a description, image, and links to the unlimited-ocr topic page so that developers can more easily learn about it.
To associate your repository with the unlimited-ocr topic, visit your repo's landing page and select "manage topics."