Hcompany/Holo-3.1-4B
Different architecture vs Qwen/Qwen3.5-0.8B. Different architecture: only 30 of 488/724 tensors match by name and shape.
7 derivatives checked
- mradermacher/Holo-3.1-4B-GGUF: Same weights, converted
- mradermacher/Holo-3.1-4B-i1-GGUF: Quantization
- prithivMLmods/Holo-3.1-4B-GGUF: Same weights, converted
- ijohn07/Holo-3.1-4B-Q4_K_M-GGUF: Quantization
- pipenetwork/Holo-3.1-4B-MLX-4bit: Quantization
- josephmayo/Holo-3.1-4B-Coder: LoRA-style (fewer layers)
- usermma/Holo-3.1-4B-abliterated-rdo: Partial fine-tune (fewer layers)
2 other repos hold byte-for-byte identical weight files.