mistralai/Ministral-3-14B-Instruct-2512-BF16
Full fine-tune vs mistralai/Ministral-3-14B-Base-2512. Fine-tune: 97% of the weights changed, across 40 of 40 layers.
Tensors changed: Embeddings 1/1, Norms 54/131, MLP 120/192, Attention 160/256, other 2/4, lm_head 1/1
14 derivatives checked
- RedHatAI/Ministral-3-14B-Instruct-2512-FP8-dynamic: Quantization
- bartowski/mistralai_Ministral-3-14B-Instruct-2512-GGUF: Same weights, converted (fewer layers)
- lmstudio-community/Ministral-3-14B-Instruct-2512-GGUF: Quantization (fewer layers)
- EnlistedGhost/Ministral-3-14B-Instruct-2512-GGUF: Same weights, converted (fewer layers)
- ggml-org/Ministral-3-14B-Instruct-2512-GGUF: Quantization (fewer layers)
- pbhappliedsystems/ministral-3-14b-instruct-2512-gguf-Q4-K-M: Quantization (fewer layers)
- mradermacher/Ministral-3-14B-Instruct-2512-BF16-i1-GGUF: Quantization (fewer layers)
- ppetermann/Ministral-3-14B-Instruct-2512-textonly-NVFP4: Quantization
- pbhappliedsystems/ministral-3-14b-instruct-2512-gguf-Q8-0: Quantization (fewer layers)
- pbhappliedsystems/ministral-3-14b-instruct-2512-gguf-Q5-K-M: Quantization (fewer layers)
- pbhappliedsystems/ministral-3-14b-instruct-2512-gguf-F16: Same weights, converted (fewer layers)
- mradermacher/Ministral-3-14B-Instruct-2512-BF16-GGUF: Quantization (fewer layers)
- RedHatAI/Ministral-3-14B-Instruct-2512-NVFP4: Different architecture
- ChetanT271/Ministral-3-14B-Instruct-MLX-4bit: Quantization