metadata

license: gemma

Gemma-3 4B Instruct GGUF Models

How to Use Gemma 3 Vision with llama.cpp

To utilize the experimental support for Gemma 3 Vision in llama.cpp, follow these steps:

Clone the lastest llama.cpp Repository:

git clone https://github.com/ggml-org/llama.cpp.git
cd llama.cpp

Build the Llama.cpp:

Build llama.cpp as usual : https://github.com/ggml-org/llama.cpp#building-the-project

Download the Gemma 3 gguf file:
Download the Gemma 3 mmproj file

Run the CLI Tool:

llama-gemma3-cli -m google_gemma-3-4b-it-f16.gguf --mmproj google_gemma-3-4b-it-mmproj-f16.gguf

Copy images to the same folder as the gguf files or alter paths appropriately

Running in chat mode, available commands: /image

> hi    
Hello! How's it going today? 

Is there something specific on your mind, or were you simply saying hi? 😊 

I’m here to chat, answer questions, help with creative tasks, or just listen – whatever you need!

> /image ./bliss.png
Encoding image ./bliss.png

> what is that
That's a beautiful image!

Choosing the Right Model Format

Selecting the correct model format depends on your hardware capabilities and memory constraints.

BF16 (Brain Float 16) – Use if BF16 acceleration is available

A 16-bit floating-point format designed for faster computation while retaining good precision.
Provides similar dynamic range as FP32 but with lower memory usage.
Recommended if your hardware supports BF16 acceleration (check your device’s specs).
Ideal for high-performance inference with reduced memory footprint compared to FP32.

📌 Use BF16 if:
✔ Your hardware has native BF16 support (e.g., newer GPUs, TPUs).
✔ You want higher precision while saving memory.
✔ You plan to requantize the model into another format.

📌 Avoid BF16 if:
❌ Your hardware does not support BF16 (it may fall back to FP32 and run slower).
❌ You need compatibility with older devices that lack BF16 optimization.

F16 (Float 16) – More widely supported than BF16

A 16-bit floating-point high precision but with less of range of values than BF16.
Works on most devices with FP16 acceleration support (including many GPUs and some CPUs).
Slightly lower numerical precision than BF16 but generally sufficient for inference.

📌 Use F16 if:
✔ Your hardware supports FP16 but not BF16.
✔ You need a balance between speed, memory usage, and accuracy.
✔ You are running on a GPU or another device optimized for FP16 computations.

📌 Avoid F16 if:
❌ Your device lacks native FP16 support (it may run slower than expected).
❌ You have memory limtations.

Quantized Models (Q4_K, Q6_K, Q8, etc.) – For CPU & Low-VRAM Inference

Quantization reduces model size and memory usage while maintaining as much accuracy as possible.

Lower-bit models (Q4_K) → Best for minimal memory usage, may have lower precision.
Higher-bit models (Q6_K, Q8_0) → Better accuracy, requires more memory.

📌 Use Quantized Models if:
✔ You are running inference on a CPU and need an optimized model.
✔ Your device has low VRAM and cannot load full-precision models.
✔ You want to reduce memory footprint while keeping reasonable accuracy.

📌 Avoid Quantized Models if:
❌ You need maximum accuracy (full-precision models are better for this).
❌ Your hardware has enough VRAM for higher-precision formats (BF16/F16).

Summary Table: Model Format Selection

Model Format	Precision	Memory Usage	Device Requirements	Best Use Case
BF16	Highest	High	BF16-supported GPU/CPUs	High-speed inference with reduced memory
F16	High	High	FP16-supported devices	GPU inference when BF16 isn’t available
Q4_K	Low	Very Low	CPU or Low-VRAM devices	Best for memory-constrained environments
Q6_K	Medium Low	Low	CPU with more memory	Better accuracy while still being quantized
Q8	Medium	Moderate	CPU or GPU with enough VRAM	Best accuracy among quantized models

Included Files & Details

`google_gemma-3-4b-it-bf16.gguf`

Model weights preserved in BF16.
Use this if you want to requantize the model into a different format.
Best if your device supports BF16 acceleration.

`google_gemma-3-4b-it-f16.gguf`

Model weights stored in F16.
Use if your device supports FP16, especially if BF16 is not available.

`google_gemma-3-4b-it-bf16-q8.gguf`

Output & embeddings remain in BF16.
All other layers quantized to Q8_0.
Use if your device supports BF16 and you want a quantized version.

`google_gemma-3-4b-it-f16-q8.gguf`

Output & embeddings remain in F16.
All other layers quantized to Q8_0.

`google_gemma-3-4b-it-q4_k_l.gguf`

Output & embeddings quantized to Q8_0.
All other layers quantized to Q4_K.
Good for CPU inference with limited memory.

`google_gemma-3-4b-it-q4_k_m.gguf`

Similar to Q4_K.
Another option for low-memory CPU inference.

`google_gemma-3-4b-it-q4_k_s.gguf`

Smallest Q4_K variant, using less memory at the cost of accuracy.
Best for very low-memory setups.

`google_gemma-3-4b-it-q6_k_l.gguf`

Output & embeddings quantized to Q8_0.
All other layers quantized to Q6_K .

`google_gemma-3-4b-it-q6_k_m.gguf`

A mid-range Q6_K quantized model for balanced performance .
Suitable for CPU-based inference with moderate memory.

`google_gemma-3-4b-it-q8.gguf`

Fully Q8 quantized model for better accuracy.
Requires more memory but offers higher precision.

Gemma 3 model card

Model Page: Gemma

Resources and Technical Documentation:

[Gemma 3 Technical Report][g3-tech-report]
[Responsible Generative AI Toolkit][rai-toolkit]
[Gemma on Kaggle][kaggle-gemma]
[Gemma on Vertex Model Garden][vertex-mg-gemma3]

Terms of Use: [Terms][terms]

Authors: Google DeepMind

Model Information

Summary description and brief definition of inputs and outputs.

Description

Gemma is a family of lightweight, state-of-the-art open models from Google, built from the same research and technology used to create the Gemini models. Gemma 3 models are multimodal, handling text and image input and generating text output, with open weights for both pre-trained variants and instruction-tuned variants. Gemma 3 has a large, 128K context window, multilingual support in over 140 languages, and is available in more sizes than previous versions. Gemma 3 models are well-suited for a variety of text generation and image understanding tasks, including question answering, summarization, and reasoning. Their relatively small size makes it possible to deploy them in environments with limited resources such as laptops, desktops or your own cloud infrastructure, democratizing access to state of the art AI models and helping foster innovation for everyone.

Inputs and outputs

Input:
- Text string, such as a question, a prompt, or a document to be summarized
- Images, normalized to 896 x 896 resolution and encoded to 256 tokens each
- Total input context of 128K tokens for the 4B, 12B, and 27B sizes, and 32K tokens for the 1B size
Output:
- Generated text in response to the input, such as an answer to a question, analysis of image content, or a summary of a document
- Total output context of 8192 tokens

Credits

Thanks Bartowski for imartix upload. And your guidance on quantization that has enabled me to produce these gguf file.