mistralai/Ministral-3-8B-Instruct-2512-BF16
Full fine-tune vs mistralai/Ministral-3-8B-Base-2512. Fine-tune: 95% of the weights changed, across 34 of 34 layers.
Tensors changed: Embeddings 1/1, Norms 27/119, MLP 102/174, Attention 136/232, other 2/4, lm_head 1/1
17 derivatives checked
- bartowski/mistralai_Ministral-3-8B-Instruct-2512-GGUF: Same weights, converted (fewer layers)
- lmstudio-community/Ministral-3-8B-Instruct-2512-GGUF: Quantization (fewer layers)
- StandardThinking/StandardOne-8B: LoRA-style
- ggml-org/Ministral-3-8B-Instruct-2512-GGUF: Quantization (fewer layers)
- mradermacher/Ministral-3-8B-Instruct-2512-BF16-i1-GGUF: Quantization (fewer layers)
- EnlistedGhost/Ministral-3-8B-Instruct-2512-GGUF: Same weights, converted (fewer layers)
- mradermacher/Ministral-3-8B-Instruct-2512-BF16-GGUF: Same weights, converted (fewer layers)
- Firworks/Ministral-3-8B-Instruct-2512-nvfp4: Quantization
- ppetermann/Ministral-3-8B-Instruct-2512-textonly-NVFP4: Quantization
- OculusMindAI/OculusMind-ToolCall-8B-v2: LoRA-style (fewer layers)
- WMTH-100d2exam/Ministral-3-8B-Instruct-2512-matura-sft-lora-distill-epod: Adapter
- keisuke-miyako/Ministral-3-8B-Instruct-2512-gguf-q4_k_m: Quantization (fewer layers)
- 0xA50C1A1/Ministral-3-8B-Instruct-2512-abliterix: Abliterated
- shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF: Quantization (fewer layers)
- IMISLab/Maistros-8B-Instruct: LoRA-style
- pankajmathur/RenCoder-Ministral-3-8B-Instruct-2512: Identical copy
- Darkhn/Ministral-3-8B-Animus-V12.0: LoRA-style