pixels & tokens @ lighton / vit.cpp, modernbert, lightonocr
AI
The launch post says the model combines text transcription, document layout, image descriptions and chart extraction.
AI
