lukaskremla/Ornith-1.5-35B-A3B-8bit-MLX
Image-Text-to-Text • 35B • Updated • 108
Complete quantized MLX conversion of Ornith 1.5 35B A3B. This collection contains vision-capable, text-only and all MTP quants.
Note Vision capable models (the ones above) contain the quantized weights and a BF 16 vision tower.
Note TextOnly models omit the vision tower to save resources.
Note MTP heads are loaded as a separate draft model in mlx-vlm or mlx-serve. Their quantization doesn't need to match the quantization of the main model.