badtheorylabs/BTL-4-Compact

Quantization vs badtheorylabs/BTL-4. badtheorylabs/BTL-4-Compact is a GGUF quantization (IQ2_XXS) of the same weights: 141 tensors match exactly and 162 within quantization error.

Tensors changed: Norms 0/101, lm_head 0/1, Embeddings 0/1, MLP 0/160, Attention 0/40

On the Hub · Downloads on Model Pulse