China Telecom open-sources the 1.2B TeleOCR model for text, layout, tables and formulas

What changed

China Telecom's XingChen team has released the 1.2B-parameter TeleOCR model for digital documents and camera-captured pages, covering text, reading order, layout elements, tables, formulas and scientific charts. Model assets, code and a technical report are available, but the published benchmark results remain vendor evaluations; teams should validate scan quality, language coverage and deployment cost on their own documents.