OpenBMB releases MiniCPM-OCR, a 1B document parser, with a 2B model that scores OCR results

OpenBMB showed MiniCPM-OCR on X on October 28, running locally on an iPhone and an iPad. It is released with open weights and can run with llama.cpp, vLLM, SGLang or Ollama; fine-tuning code is open source, and OpenBMB also offers a demo and an API.

Two models are described:

• MiniCPM-OCR 1.0, a document parser at the 1B scale. OpenBMB claims state-of-the-art OCR performance and speed at that size, with 93.68 on Real5-OmniDocBench and 54.7 on OCRBench v2. It is shown on handwritten notes, complex formulas, newspapers and poor-quality scans.
• MiniCPM-OCR Jev, a 2B decision model that checks OCR output. It compares parsed text, formulas and tables with the original image and gives each a correctness score from 0 to 1, so users can pick better results and see what needs human review. OpenBMB says it approaches Qwen3.5-397B-A17B on JevOCRBench and works with the output of different OCR parsers.

The benchmark figures are OpenBMB's own claims, not independently measured.