mradermacher/Pristine-Mini-8B-A1B-Base-GGUF

Same weights, converted vs qingy2024/Pristine-Mini-8B-A1B-Base. Same weights stored in another dtype (BF16 → F16/F32), plus 76 new tensors.

Tensors changed: Attention 0/20, Norms 0/81, MLP 0/79, Embeddings 0/1, lm_head 0/1

On the Hub · Downloads on Model Pulse