---

language:

• en
• multilingual

tags:

• mistral
• ministral
• reasoning
• mxfp4
• gguf
• vision
• multimodal
• coding

license: apache-2.0

base_model: mistralai/Ministral-3-8B-Reasoning-2512

---

Ministral 3 8B Reasoning-2512 — MXFP4 GGUF

> Note on Thinking/Reasoning Display: The model thinks correctly (generates [THINK]...[/THINK] tags internally), but LM Studio 0.4.17 does not display thinking in a collapsible reasoning section. LM Studio only natively supports DeepSeek-style <think> tags for reasoning display. The model works perfectly via llama-server with --reasoning on flag. This is a known limitation of LM Studio for the Ministral reasoning format.

MXFP4 quantization of mistralai/Ministral-3-8B-Reasoning-2512, an 8B parameter reasoning model from Mistral AI.

About the Model

Ministral 3 8B Reasoning is a dense transformer with 40 layers, 4096 hidden dimension, and 24-layer Pixtral ViT vision encoder. It supports:

• Advanced reasoning with chain-of-thought capabilities
• Code generation and debugging across multiple languages
• Vision understanding via multimodal image input
• Tool calling with native function calling support
• 262K context window with YaRN scaling

Quantization

This GGUF was quantized from Mistral's official BF16 GGUF using llama.cpp (build 537). The BF16 weights were dequantized to F32 during quantization, then quantized to MXFP4 format.

MXFP4 (Microscaling FP4) uses block-wise quantization with shared exponents per block.

Files

| File | Size | Description |

|------|------|-------------|

| ministral-3-8b-reasoning-2512-mxfp4.gguf | ~3.5 GB | MXFP4 quantized model weights |

| mmproj-ministral-3-8b-reasoning-2512-f16.gguf | ~693 MB | Vision projector (F16, unquantized) |

Usage

llama.cpp

# Server mode with OpenAI-compatible API
llama-server \
  -m ministral-3-8b-reasoning-2512-mxfp4.gguf \
  --mmproj mmproj-ministral-3-8b-reasoning-2512-f16.gguf \
  -ngl 99 \
  --host 0.0.0.0 \
  --port 8080

LM Studio

1. Download both files from this repository
2. Load the main GGUF file in LM Studio
3. Load the mmproj file for vision support
4. Set GPU offload layers to maximum

Architecture

• Parameters: 8B (dense transformer)
• Layers: 40
• Hidden dimension: 4096
• Attention heads: 32 (8 KV heads for GQA)
• Vision encoder: 24-layer Pixtral ViT
• Context: 262K (YaRN scaled)
• Vocabulary: Mistral Tekken tokenizer

Hardware Requirements

• Minimum: 6 GB VRAM for text-only
• Recommended: 8 GB VRAM for full GPU offload

License

Apache 2.0 — same as the base model.