PaddlePaddle/PaddleOCR-VL-1.6
Image-Text-to-Text • 1.0B • Updated • 9.92k • 278
https://huggingface.co/google/gemma-4-31B-it
I would recommend this model. very very good for its size. achieves same/better results than models 10x its size.
Pair that with hermes-agent by nous research which takes your model and makes it an agent to do continuous tasks and automated things. and you got a pretty good setup.