Baidu has open-sourced Unlimited-OCR, a 3-billion-parameter document parser that reads entire 100-page PDFs in one shot—without splitting them into chunks or losing context across pages. Released on June 22, 2026, the model targets a persistent weakness in document AI: the tendency of conventional pipelines to fragment long documents page by page, breaking tables, cross-references, and reading order while accumulating errors deep into a file.
Continue reading
The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in✓ Signed in — this article isn’t included in your current plan.