FancieF/Qwythos-27B-v1-MLX-VLM-4bit
Image-Text-to-Text • 5B • Updated • 104
Qwythos-27B MLX
Apple Silicon MLX conversions of empero-ai/Qwythos-27B-v1: VLM 4-bit (default), VLM BF16, and MTP drafter. Use with mlx-vlm.
Note Default for ~32GB Macs. Vision kept. Pair with MTP for speed.
FancieF/Qwythos-27B-v1-MLX-MTP
0.4B • Updated • 95Note Not a full LLM. Use as --draft-model with mlx-vlm.
FancieF/Qwythos-27B-v1-MLX-VLM-bf16
Image-Text-to-Text • 27B • Updated • 225Note ~55GB BF16. High RAM / re-quant. Vision kept.