microsoft/Fara1.5-27B
Full fine-tune vs Qwen/Qwen3.5-27B. Fine-tune: 95% of the weights changed, across 64 of 64 layers.
Tensors changed: lm_head 1/1, Embeddings 3/4, MLP 303/304, Attention 524/556, Norms 259/319
6 derivatives checked
- bartowski/Fara1.5-27B-GGUF: Same weights, converted
- prithivMLmods/Fara1.5-27B-GGUF: Same weights, converted
- Abiray/Fara1.5-27B-GGUF: Quantization
- mlx-community/Fara1.5-27B-OptiQ-4bit: Quantization
- DevQuasar/microsoft.Fara1.5-27B-GGUF: Quantization
- prithivMLmods/Fara1.5-27B-FP8: Quantization
2 other repos hold byte-for-byte identical weight files.