mistralai/Shieldstral-1.0-3B
Full fine-tune vs mistralai/Ministral-3-3B-Base-2512. Fine-tune: 92% of the weights changed, across 26 of 26 layers.
Tensors changed: Embeddings 1/1, Norms 20/103, MLP 93/150, Attention 124/200, other 2/4
6 derivatives checked
- noctrex/Shieldstral-1.0-3B-GGUF: Same weights, converted (fewer layers)
- vroomfondel/Shieldstral-1.0-3B-NVFP4-ModelOpt: Quantization
- Abiray/Shieldstral-1.0-3B-GGUF: Quantization (fewer layers)
- NANI-Nithin/Shieldstral-1.0-3B-GGUF: Same weights, converted (fewer layers)
- Metabaron6/Shieldstral-1.0-3B-GGUF: Same weights, converted (fewer layers)
- AXONVERTEX-AI-RESEARCH/Shieldstral-1.0-3B-MLX-4bit: Quantization