Open-source mistral-community_pixtral-12b-GGUF model - Supports image-text to text conversion tasks

Mistral Community Pixtral 12b GGUF

Developed by bartowski

This is the quantized version of the pixtral-12b model, quantized using llama.cpp, supporting image-text-to-text tasks.

Open Source License:Apache-2.0 #Image Text Generation #Multimodal Quantization #Efficient Inference

Downloads 1,728

Release Time : 4/25/2025

Model Overview

This model is the quantized version of pixtral-12b, suitable for image-text-to-text tasks, supporting multiple quantization types and ideal for running on resource-limited devices.

Model Features

Multiple Quantization Options

Offers various quantization types from BF16 to Q2_K to meet different hardware and performance needs.

Supports Image-Text-to-Text

Capable of handling image-text-to-text tasks, suitable for multimodal application scenarios.

Optimized Hardware Compatibility

Supports running on platforms like LM Studio and llama.cpp, with optimized inference performance for ARM and AVX CPUs.

Model Capabilities

Image Text Generation

Multimodal Processing

Efficient Inference

Use Cases

Multimodal Applications

Image Caption Generation

Generates detailed textual descriptions based on input images.

Visual Question Answering

Answers natural language questions about image content.

🚀 Llamacpp imatrix Quantizations of pixtral-12b by mistral-community

This project provides quantized versions of the pixtral-12b model by the mistral-community. It uses llama.cpp for quantization, enabling users to run the model more efficiently on various hardware.

Metadata

Property	Details
Quantized By	bartowski
Pipeline Tag	image-text-to-text
Base Model	mistral-community/pixtral-12b
Base Model Relation	quantized
License	apache-2.0

🚀 Quick Start

Quantization: Using llama.cpp release b5173 for quantization.
Original Model: You can find the original model here.
Running the Model:
- Run the quantized models in LM Studio.
- Run them directly with llama.cpp, or any other llama.cpp based project.

✨ Features

Prompt Format

No prompt format found, check the original model page.

What's New

Fixed mmproj files with fixes from the mainline: PR 13208.

📦 Installation

Download a File

You can download a specific file from the following table:

Filename	Quant type	File Size	Split	Description
pixtral-12b-bf16.gguf	bf16	24.50GB	false	Full BF16 weights.
pixtral-12b-Q8_0.gguf	Q8_0	13.02GB	false	Extremely high quality, generally unneeded but max available quant.
pixtral-12b-Q6_K_L.gguf	Q6_K_L	10.38GB	false	Uses Q8_0 for embed and output weights. Very high quality, near perfect, recommended.
pixtral-12b-Q6_K.gguf	Q6_K	10.06GB	false	Very high quality, near perfect, recommended.
pixtral-12b-Q5_K_L.gguf	Q5_K_L	9.14GB	false	Uses Q8_0 for embed and output weights. High quality, recommended.
pixtral-12b-Q5_K_M.gguf	Q5_K_M	8.73GB	false	High quality, recommended.
pixtral-12b-Q5_K_S.gguf	Q5_K_S	8.52GB	false	High quality, recommended.
pixtral-12b-Q4_K_L.gguf	Q4_K_L	7.98GB	false	Uses Q8_0 for embed and output weights. Good quality, recommended.
pixtral-12b-Q4_1.gguf	Q4_1	7.80GB	false	Legacy format, similar performance to Q4_K_S but with improved tokens/watt on Apple silicon.
pixtral-12b-Q4_K_M.gguf	Q4_K_M	7.48GB	false	Good quality, default size for most use cases, recommended.
pixtral-12b-Q3_K_XL.gguf	Q3_K_XL	7.15GB	false	Uses Q8_0 for embed and output weights. Lower quality but usable, good for low RAM availability.
pixtral-12b-Q4_K_S.gguf	Q4_K_S	7.12GB	false	Slightly lower quality with more space savings, recommended.
pixtral-12b-IQ4_NL.gguf	IQ4_NL	7.10GB	false	Similar to IQ4_XS, but slightly larger. Offers online repacking for ARM CPU inference.
pixtral-12b-Q4_0.gguf	Q4_0	7.09GB	false	Legacy format, offers online repacking for ARM and AVX CPU inference.
pixtral-12b-IQ4_XS.gguf	IQ4_XS	6.74GB	false	Decent quality, smaller than Q4_K_S with similar performance, recommended.
pixtral-12b-Q3_K_L.gguf	Q3_K_L	6.56GB	false	Lower quality but usable, good for low RAM availability.
pixtral-12b-Q3_K_M.gguf	Q3_K_M	6.08GB	false	Low quality.
pixtral-12b-IQ3_M.gguf	IQ3_M	5.72GB	false	Medium-low quality, new method with decent performance comparable to Q3_K_M.
pixtral-12b-Q3_K_S.gguf	Q3_K_S	5.53GB	false	Low quality, not recommended.
pixtral-12b-Q2_K_L.gguf	Q2_K_L	5.45GB	false	Uses Q8_0 for embed and output weights. Very low quality but surprisingly usable.
pixtral-12b-IQ3_XS.gguf	IQ3_XS	5.31GB	false	Lower quality, new method with decent performance, slightly better than Q3_K_S.
pixtral-12b-IQ3_XXS.gguf	IQ3_XXS	4.95GB	false	Lower quality, new method with decent performance, comparable to Q3 quants.
pixtral-12b-Q2_K.gguf	Q2_K	4.79GB	false	Very low quality but surprisingly usable.
pixtral-12b-IQ2_M.gguf	IQ2_M	4.44GB	false	Relatively low quality, uses SOTA techniques to be surprisingly usable.
pixtral-12b-IQ2_S.gguf	IQ2_S	4.14GB	false	Low quality, uses SOTA techniques to be usable.

Downloading using huggingface-cli

Click to view download instructions

First, make sure you have huggingface-cli installed:

pip install -U "huggingface_hub[cli]"

Then, you can target the specific file you want:

huggingface-cli download bartowski/mistral-community_pixtral-12b-GGUF --include "mistral-community_pixtral-12b-Q4_K_M.gguf" --local-dir ./

If the model is bigger than 50GB, it will have been split into multiple files. In order to download them all to a local folder, run:

huggingface-cli download bartowski/mistral-community_pixtral-12b-GGUF --include "mistral-community_pixtral-12b-Q8_0/*" --local-dir ./

You can either specify a new local-dir (mistral-community_pixtral-12b-Q8_0) or download them all in place (./).

💡 Usage Tips

Embed/output weights

Some of these quants (e.g., Q3_K_XL, Q4_K_L) use the standard quantization method with the embeddings and output weights quantized to Q8_0 instead of the normal default.

ARM/AVX information

Previously, you would download Q4_0_4_4/4_8/8_8, and their weights would be interleaved in memory to improve performance on ARM and AVX machines by loading more data in one pass.

Now, there is "online repacking" for weights. Details can be found in this PR. If you use Q4_0 and your hardware would benefit from repacking weights, it will do it automatically on the fly.

As of llama.cpp build b4282, you will not be able to run the Q4_0_X_X files and will instead need to use Q4_0.

Additionally, if you want slightly better quality, you can use IQ4_NL thanks to this PR, which will also repack the weights for ARM (only the 4_4 for now). The loading time may be slower, but it will result in an overall speed increase.

Click to view Q4_0_X_X information (deprecated)

I'm keeping this section to show the potential theoretical uplift in performance from using the Q4_0 with online repacking.

Click to view benchmarks on an AVX2 system (EPYC7702)

model	size	params	backend	threads	test	t/s	% (vs Q4_0)
qwen2 3B Q4_0	1.70 GiB	3.09 B	CPU	64	pp512	204.03 ± 1.03	100%
qwen2 3B Q4_0	1.70 GiB	3.09 B	CPU	64	pp1024	282.92 ± 0.19	100%
qwen2 3B Q4_0	1.70 GiB	3.09 B	CPU	64	pp2048	259.49 ± 0.44	100%
qwen2 3B Q4_0	1.70 GiB	3.09 B	CPU	64	tg128	39.12 ± 0.27	100%
qwen2 3B Q4_0	1.70 GiB	3.09 B	CPU	64	tg256	39.31 ± 0.69	100%
qwen2 3B Q4_0	1.70 GiB	3.09 B	CPU	64	tg512	40.52 ± 0.03	100%
qwen2 3B Q4_K_M	1.79 GiB	3.09 B	CPU	64	pp512	301.02 ± 1.74	147%
qwen2 3B Q4_K_M	1.79 GiB	3.09 B	CPU	64	pp1024	287.23 ± 0.20	101%
qwen2 3B Q4_K_M	1.79 GiB	3.09 B	CPU	64	pp2048	262.77 ± 1.81	101%
qwen2 3B Q4_K_M	1.79 GiB	3.09 B	CPU	64	tg128	18.80 ± 0.99	48%
qwen2 3B Q4_K_M	1.79 GiB	3.09 B	CPU	64	tg256	24.46 ± 3.04	83%
qwen2 3B Q4_K_M	1.79 GiB	3.09 B	CPU	64	tg512	36.32 ± 3.59	90%
qwen2 3B Q4_0_8_8	1.69 GiB	3.09 B	CPU	64	pp512	271.71 ± 3.53	133%
qwen2 3B Q4_0_8_8	1.69 GiB	3.09 B	CPU	64	pp1024	279.86 ± 45.63	100%
qwen2 3B Q4_0_8_8	1.69 GiB	3.09 B	CPU	64	pp2048	320.77 ± 5.00	124%
qwen2 3B Q4_0_8_8	1.69 GiB	3.09 B	CPU	64	tg128	43.51 ± 0.05	111%
qwen2 3B Q4_0_8_8	1.69 GiB	3.09 B	CPU	64	tg256	43.35 ± 0.09	110%
qwen2 3B Q4_0_8_8	1.69 GiB	3.09 B	CPU	64	tg512	42.60 ± 0.31	105%

Q4_0_8_8 offers a nice bump to prompt processing and a small bump to text generation.

Which file should I choose?

Click here for details

A great write-up with charts showing various performances is provided by Artefact2 here.

The first thing to figure out is how big a model you can run. To do this, you'll need to figure out how much RAM and/or VRAM you have.

Fastest Performance: If you want your model running as fast as possible, aim to fit the whole thing on your GPU's VRAM. Choose a quant with a file size 1 - 2GB smaller than your GPU's total VRAM.
Maximum Quality: If you want the absolute maximum quality, add both your system RAM and your GPU's VRAM together, then choose a quant with a file size 1 - 2GB smaller than that total.

📄 License

This project is licensed under the Apache 2.0 License.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご