I just noticed this new LightOnOCR-2-1B model, they claim it is very good, but is it better then deepseek-ocr, or does deepseek ocr come with a newer version soon? A lot changed in the ai world to optimize and make models smaller and faster (like iquants).
Speed is an important factor when you need to scan many pages of a pdf. On the other hand multi-lingual understand allows models to be self-corrective even in other languages then english.
I just noticed this new LightOnOCR-2-1B model, they claim it is very good, but is it better then deepseek-ocr, or does deepseek ocr come with a newer version soon? A lot changed in the ai world to optimize and make models smaller and faster (like iquants).
Speed is an important factor when you need to scan many pages of a pdf. On the other hand multi-lingual understand allows models to be self-corrective even in other languages then english.