An 8-bit, group-size-64 MLX quantization of PaddlePaddle/PaddleOCR-VL-1.5, for on-device OCR on Apple Silicon. The newer PaddleOCR-VL-1.6 is an architecture-identical drop-in successor — this 1.5 quant is kept for reference.


Base model	`PaddlePaddle/PaddleOCR-VL-1.5`
Quantization	8-bit, group size 64, affine
Format	MLX `safetensors`
Size	~1.0 GB

License & attribution

Apache-2.0, inherited from the base model. All credit for the model goes to the PaddlePaddle / PaddleOCR team — this repository only provides an MLX-quantized copy.

translate-studio/PaddleOCR-VL-1.5-8bit

PaddleOCR-VL-1.5 — 8-bit (MLX)

License & attribution

translate-studio/
PaddleOCR-VL-1.5-8bit