mistralai/Pixtral-12B-2409
Full fine-tune vs mistralai/Pixtral-12B-Base-2409. Fine-tune: 100% of the weights changed, across 40 of 40 layers.
Tensors changed: Attention 256/256, Norms 105/130, MLP 192/192, lm_head 1/1, Embeddings 1/1, other 5/5
6 derivatives checked
- unsloth/Pixtral-12B-2409: Same weights, repacked
- EnlistedGhost/Pixtral-12B-2409-GGUF: Same weights, converted (fewer layers)
- unsloth/Pixtral-12B-2409-bnb-4bit: Quantization
- unsloth/Pixtral-12B-2409-unsloth-bnb-4bit: Quantization
- unalignment/Pixtral-12B-Captioner-Relaxed: Full fine-tune
- ServiceNow/Pixtral-12B-2409-StarFlow: Full fine-tune
7 other repos hold byte-for-byte identical weight files.