metadata
base_model:
- microsoft/Phi-4-mini-reasoning
- microsoft/Phi-4-mini-instruct
license: mit
tags:
- merge
pipeline_tag: text-generation
library_name: transformers
This is an upscaled/merged version of microsoft/Phi-4-mini-instruct. it has 6B parameters and 56 layers. As it has more layers, it also has an increased capacity for learning, so this model can be fine-tuned to be, very good.
Layers 18-22 are merged from microsoft/Phi-4-mini-reasoning, others are from microsoft/Phi-4-mini-instruct.