OpenBMB ships 1.3B MiniCPM-OCR 1.0 plus a 2B verifier model, Apache 2.0
BerryBlush_Rogers691 · reddit · 2026-10-28
OpenBMB released MiniCPM-OCR 1.0 (1.3B, Chinese/English document parsing) alongside MiniCPM-OCR Jev (2B), which scores OCR output quality 0-1 from an image crop, handling text, formulas and tables — and can also grade results from other OCR parsers.
A benchmark chart shows MiniCPM-OCR 1.0 topping open-source OCR models (91.3 average across OmniDocBench v1.6, Real5, Wild-OmniDocBench, olmOCR-bench) at 1.3 pages/s, ahead of Xiaomi-OCR-0 and PaddleOCR-VL-1.6. Code and weights are Apache 2.0, with local deployment instructions and quantized weights on GitHub. The poster suggests using it in local RAG pipelines to flag questionable regions before indexing.
Related event: OpenBMB Open-Sources MiniCPM-OCR Family with 1.3B SOTA Model(6 posts)→
More from Models
- Liquid AI launches LFM3 family: 6 compact models plus d1 decision models in a single pass — antoine_chaffin · 2026-10-28
- Follow-up: GPT-5.6 terra would be a hit if OpenAI cut its price 80% instead of 20% — JoshPurtell · 2026-10-28
- Dev calls out OpenAI's messy GPT-5.6 pricing: mini/nano hiked, luna reverted, terra left behind — JoshPurtell · 2026-10-28
- Liquid AI LFM3's hybrid architecture: 6 global attention layers, 256K context, 25% faster inference — cneuralnetwork · 2026-10-28
- Dev names GLM 5.5 Flash and Qwen 4 27B as the models he's most excited about — victormustar · 2026-10-28
- Maxime Labonne demos d1: bounding boxes for named objects in a single forward pass — maximelabonne · 2026-10-28