OpenBMB open-sources MiniCPM-OCR: 1.3B model parses 128 pages per minute
haider1 · x · 2026-10-31
OpenBMB released MiniCPM-OCR, an open-source (Apache-2.0) OCR model family with two components. MiniCPM-OCR 1.0 (1.3B) parses Chinese and English documents and PDFs, handling scans, photos, curved pages, and recaptured documents at up to 128 pages per minute with its optimized SDK. MiniCPM-OCR Jev (2B) evaluates OCR results at the region level, flagging unreliable regions for review and working alongside different parsing models. The author's test—running OCR through 1.0 and verifying accuracy with Jev—showed impressive results for models this size, supporting the thesis that small specialized models will take over tasks once requiring much larger systems. Free API, web demo, and local deployment are available; the repo has 640 stars.
Related event: OpenBMB Releases MiniCPM-OCR: Small Models Punch Above Their Weight(2 posts)→
More from Models
- ThursdAI crew debates: with Opus 7 coming, who still cares about better benchmarks? — thursdai_pod · 2026-10-31
- 14 models tested on 960 analyst tasks: AI scores 97/100 as analyst, picks winners no better than random — HeyAmit_ · 2026-10-31
- Anthropic adds invisible text watermark to Claude outputs starting Nov 3 — petesena · 2026-10-31
- GPT-6 Luna vs GPT-6.1 Terra vs GPT-6.1 Sol vs GPT-6 Astra in Pixel Art — Angaisb_ · 2026-10-31
- Perplexity engineer invites public tasks to prove or fix pplx-decider's SOTA claims — antoine_chaffin · 2026-10-31
- Arena weekly: Gemini 4 Carbon tops Text Arena, GPT-6.1 Terra enters Agent Arena at #8 — arena · 2026-10-31