blascotobasco/L3.2-Ascendant-Prime-16E-A6B
MoE built from dense models vs ChiKoi7/GPT-5-Distill-llama3.2-3B-Instruct-Heretic-GGUF. blascotobasco/L3.2-Ascendant-Prime-16E-A6B is a mixture of 16 experts assembled from dense models (mergekit-moe style): expert 13 is an exact copy of ChiKoi7/GPT-5-Distill-llama3.2-3B-Instruct-Heretic-GGUF's MLP; the other experts come from other models. Shared tensors (attention, norms, embeddings): 58 of 170 identical.
3 derivatives checked
- mradermacher/L3.2-Ascendant-Prime-16E-A6B-i1-GGUF: Quantization
- mradermacher/L3.2-Ascendant-Prime-16E-A6B-GGUF: Quantization
- blascotobasco/L3.2-Ascendant-Prime-16E-A6B-GGUF: Same weights, converted