microsoft/Fara1.5-9B
Partial fine-tune vs Qwen/Qwen3.5-9B. Partial change: 89% of the weights changed (attention, embeddings, lm_head, mlp, norms), in 32 of 32 layers.
Tensors changed: lm_head 1/1, Embeddings 3/4, MLP 207/208, Attention 308/332, Norms 176/215
7 derivatives checked
- bartowski/Fara1.5-9B-GGUF: Same weights, converted
- webmp3/Sakura-Fara1.5-9B-GGUF: Quantization
- prithivMLmods/Fara1.5-9B-GGUF: Same weights, converted
- mlx-community/Fara1.5-9B-OptiQ-4bit: Quantization
- DevQuasar/microsoft.Fara1.5-9B-GGUF: Quantization
- WootzappLab/fara-9b-baseline-eval: Identical copy
- Tdamre/Fara1.5-9B-GGUF: Quantization
2 other repos hold byte-for-byte identical weight files.