This is a big step forward for 3B class models. Trained on smol PIPPA, alpaca-cleaned, and two custom datasets, and based on https://huggingface.co/jeiku/Rosa_v3_3B

This should serve as a decent fiction model, though it also excels at roleplaying, but is not an ideal model for logical queries or riddles.

Downloads last month
20
Safetensors
Model size
2.8B params
Tensor type
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Datasets used to train jeiku/Foundation_3B