OpenBMB releases MiniCPM-OCR 1.0, a 1.3B document model
Original title: openbmb/MiniCPM-OCR-1.0 — new model trending #30 on Hugging Face
The publisher reports 93.68 on Real5-OmniDocBench and 1.2 pages per second on an RTX 5090.
OpenBMB published MiniCPM-OCR 1.0, a 1.3-billion-parameter OCR model in the MiniCPM-V/o family for text, tables, formulas, charts, seals and structured extraction from Chinese and English documents. It claims leading scores on OmniDocBench v1.6, Wild-OmniDocBench and olmOCR-Bench, plus 54.7 on OCRBench v2. DSpark acceleration is said to raise end-to-end efficiency by 55%, with day-one support for vLLM, SGLang, llama.cpp and Ollama.
Unconfirmed: Benchmark scores and throughput figures come from OpenBMB’s own model card and are not independently verified here.
- Throughput is claimed at about 1.49× PaddleOCR-VL-1.6 and 3.62× MinerU2.5 Pro.
- The RTX 5090 test used all 1,651 OmniDocBench v1.6 images at image-level concurrency 1.
- Quantized variants, a pip SDK, an API and agent workflows are included.
- The card was trending at No. 30 on Hugging Face.
- Vendors
- OpenBMB
- Products
- DSpark
- AI models
- MiniCPM-OCR 1.0
From the original article
5 min read at huggingface.coimage-text-to-text · 791 downloads · 215 likes · license apache-2. 0
By openbmb · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.