Ministral-3-14B-Reasoning-2512 needs about 10.1 GB of memory at Q4_K_M (a 8.2 GB file) with an 8K context window, so it fits on a 12 GB GPU. This repository has 26 quants, from 3.4 GB (IQ1_S) to 27.0 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 5.4 GB.
Ministral‑3‑14B‑Reasoning‑2512: 14B LLM with vision and advanced reasoning
Ministral‑3‑14B‑Reasoning‑2512 is a 14 billion‑parameter model in the Mistral family. It includes a 13.5 B language core and a 0.4 B vision encoder. The model supports a 256k token context window and is trained for complex reasoning tasks.
Good at
- Advanced reasoning and math
- Vision‑enabled image analysis
- Multilingual support
Limits
- Large 256k context requires high memory
- No dedicated GPU acceleration beyond llama.cpp
- Requires Apache‑2.0 license compliance
Best for
- Local AI assistants with reasoning needs
- Math and coding problem solving
- Vision‑based data extraction
Who it’s for. Developers and researchers who need a powerful reasoning model that can run on edge hardware. The model is suitable for private deployments where advanced capabilities must fit within 32 GB VRAM in BF16 or less than 24 GB when quantized.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ1_S | 3.4 GB | 5.3 GB | Fits | Get file → | |
| IQ1_M | 3.7 GB | 5.5 GB | Fits | Get file → | |
| IQ2_XXS | 4.1 GB | 5.9 GB | Fits | Get file → | |
| IQ2_M | 4.9 GB | 6.8 GB | Fits | Get file → | |
| Q2_K | 5.2 GB | 7.1 GB | Fits | Get file → | |
| Q2_K_L | 5.4 GB | 7.2 GB | Tight | Get file → | |
| IQ3_XXS | 5.5 GB | 7.3 GB | Tight | Get file → | |
| Q2_K_XL | 5.5 GB | 7.4 GB | Tight | Get file → | |
| Q3_K_S | 6.1 GB | 7.9 GB | Tight | Get file → | |
| Q3_K_M | 6.7 GB | 8.5 GB | Too big | Get file → | |
| Q3_K_XL | 6.9 GB | 8.8 GB | Too big | Get file → | |
| IQ4_XS | 7.4 GB | 9.3 GB | Too big | Get file → | |
| IQ4_NL | 7.8 GB | 9.6 GB | Too big | Get file → | |
| Q4_0 | 7.8 GB | 9.6 GB | Too big | Get file → | |
| Q4_K_S | 7.8 GB | 9.7 GB | Too big | Get file → | |
| Q4_K_M | 8.2 GB | 10.1 GB | Too big | Get file → | |
| Q4_K_XL | 8.4 GB | 10.2 GB | Too big | Get file → | |
| Q4_1 | 8.6 GB | 10.4 GB | Too big | Get file → | |
| Q5_K_S | 9.4 GB | 11.2 GB | Too big | Get file → | |
| Q5_K_M | 9.6 GB | 11.5 GB | Too big | Get file → | |
| Q5_K_XL | 9.6 GB | 11.5 GB | Too big | Get file → | |
| Q6_K | 11.1 GB | 12.9 GB | Too big | Get file → | |
| Q6_K_XL | 12.1 GB | 14.0 GB | Too big | Get file → | |
| Q8_0 | 14.4 GB | 16.2 GB | Too big | Get file → | |
| Q8_K_XL | 17.1 GB | 19.0 GB | Too big | Get file → | |
| BF16 | 27.0 GB | 28.9 GB | Too big | Get file → |
For 8 GB VRAM use Q4_K_S (7.83 GB file, ~9.7 GB memory). For 12 GB VRAM use Q4_K_M (8.24 GB file, ~10.1 GB memory). For 16 GB VRAM use Q4_K_XL (8.37 GB file, ~10.2 GB memory). Memory estimates are approximate.
What GPU do you need for Ministral-3-14B-Reasoning-2512?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: no quant fits at an 8K context.
- 6 GB VRAM: the best quant that fits is IQ2_XXS (4.1 GB file, about 5.9 GB in use).
- 8 GB VRAM: the best quant that fits is Q3_K_S (6.1 GB file, about 7.9 GB in use).
- 12 GB VRAM: the best quant that fits is Q5_K_XL (9.6 GB file, about 11.5 GB in use).
- 16 GB VRAM: the best quant that fits is Q6_K_XL (12.1 GB file, about 14.0 GB in use).
- 24 GB VRAM: the best quant that fits is Q8_K_XL (17.1 GB file, about 19.0 GB in use).
- 32 GB VRAM: the best quant that fits is BF16 (27.0 GB file, about 28.9 GB in use).
Which Ministral-3-14B-Reasoning-2512 quant should you choose?
Q4_K_M is the usual starting point for Ministral-3-14B-Reasoning-2512: it balances file size and quality well. Here it is a 8.2 GB file.
The Q8_0 file is 14.4 GB, about 1.7 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.
If memory is tight, the smallest file is IQ1_S at 3.4 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Ministral-3-14B-Reasoning-2512
How much VRAM does Ministral-3-14B-Reasoning-2512 need?
About 10.1 GB at Q4_K_M with an 8K context window. That is the 8.2 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run Ministral-3-14B-Reasoning-2512 on an 8 GB GPU?
Yes, with Q3_K_S: the 6.1 GB file needs about 7.9 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Ministral-3-14B-Reasoning-2512 GGUF?
It depends on the quant: IQ1_S is 3.4 GB, IQ1_M is 3.7 GB, IQ2_XXS is 4.1 GB, IQ2_M is 4.9 GB, Q2_K is 5.2 GB, Q2_K_L is 5.4 GB and 20 more.
What context length does Ministral-3-14B-Reasoning-2512 support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 1.3 GB at 8K and 5.4 GB at 32K.
Where can I download Ministral-3-14B-Reasoning-2512 GGUF?
The files are on Hugging Face in the unsloth/Ministral-3-14B-Reasoning-2512-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does Ministral-3-14B-Reasoning-2512 use?
Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run this model on 8 GB VRAM?
Yes, using Q4_K_S quantization the model file is 7.83 GB and requires about 9.7 GB VRAM at 8k context.
Does the model support vision tasks?
Yes, it includes a 0.4 B vision encoder for image analysis.
What license applies to this model?
The model is released under the Apache‑2.0 license, allowing commercial and non‑commercial use.