Mozilla packaged the Gemma 3 models into executable weights that we
call llamafiles. This gives
you the easiest fastest way to use the model on Linux, MacOS, Windows,
FreeBSD, OpenBSD and NetBSD systems you control on both AMD64 and ARM64.
Software Last Updated: 2025-03-31
Llamafile Version: 0.9.2
Quickstart
To get started, you need both the Gemma 3 weights, and the llamafile
software. Both of them are included in a single file, which can be
downloaded and run as follows:
The default mode of operation for these llamafiles is our new command
line chatbot interface.
Usage
You can use triple quotes to ask questions on multiple lines. You can
pass commands like /stats and /context to see runtime status
information. You can change the system prompt by passing the -p "new system prompt" flag. You can press CTRL-C to interrupt the model.
Finally CTRL-D may be used to exit.
If you prefer to use a web GUI, then a --server mode is provided, that
will open a tab with a chatbot and completion interface in your browser.
For additional help on how it may be used, pass the --help flag. The
server also has an OpenAI API compatible completions endpoint that can
be accessed via Python using the openai pip package.
./google_gemma-3-12b-it-Q6_K.llamafile --server
An advanced CLI mode is provided that's useful for shell scripting. You
can use it by passing the --cli flag. For additional help on how it
may be used, pass the --help flag.
./google_gemma-3-12b-it-Q6_K.llamafile --cli -p 'four score and seven' --log-disable
On Windows there's a 4GB limit on executable sizes.
Context Window
This model has a max context window size of 128k tokens. By default, a
context window size of 8192 tokens is used. You can ask llamafile
to use the maximum context size by passing the -c 0 flag. That's big
enough for a small book. If you want to be able to have a conversation
with your book, you can use the -f book.txt flag.
GPU Acceleration
On GPUs with sufficient RAM, the -ngl 999 flag may be passed to use
the system's NVIDIA or AMD GPU(s). On Windows, only the graphics card
driver needs to be installed if you own an NVIDIA GPU. On Windows, if
you have an AMD GPU, you should install the ROCm SDK v6.1 and then pass
the flags --recompile --gpu amd the first time you run your llamafile.
On NVIDIA GPUs, by default, the prebuilt tinyBLAS library is used to
perform matrix multiplications. This is open source software, but it
doesn't go as fast as closed source cuBLAS. If you have the CUDA SDK
installed on your system, then you can pass the --recompile flag to
build a GGML CUDA library just for your system that uses cuBLAS. This
ensures you get maximum performance.
llamafile is a new format introduced by Mozilla on Nov 20th 2023. It
uses Cosmopolitan Libc to turn LLM weights into runnable llama.cpp
binaries that run on the stock installs of six OSes for both ARM64 and
AMD64.
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
Usage
Below, there are some code snippets on how to get quickly started with running the model. First, install the Transformers library. Gemma 3 is supported starting from transformers 4.50.0.
$ pip install -U transformers
Then, copy the snippet from the section that is relevant for your use case.
Running with the pipeline API
With instruction-tuned models, you need to use chat templates to process our inputs first. Then, you can pass it to the pipeline.
python
1from transformers import pipeline
23pipe = pipeline("text-generation", model="google/gemma-3-12b-it", device="cuda", torch_dtype=torch.bfloat16)45messages =[6[7{8"role":"system",9"content":[{"type":"text","text":"You are a helpful assistant."},]10},11{12"role":"user",13"content":[{"type":"text","text":"Write a poem on Hugging Face, the company"},]14},15],16]1718output = pipe(messages, max_new_tokens=50)
Running the model on a single / multi GPU
python
1from transformers import AutoTokenizer, BitsAndBytesConfig, Gemma3ForCausalLM
2import torch
34model_id ="google/gemma-3-12b-it"56quantization_config = BitsAndBytesConfig(load_in_8bit=True)78model = Gemma3ForCausalLM.from_pretrained(9 model_id, quantization_config=quantization_config
10).eval()1112tokenizer = AutoTokenizer.from_pretrained(model_id)1314messages =[15[16{17"role":"system",18"content":[{"type":"text","text":"You are a helpful assistant."},]19},20{21"role":"user",22"content":[{"type":"text","text":"Write a poem on Hugging Face, the company"},]23},24],25]26inputs = tokenizer.apply_chat_template(27 messages,28 add_generation_prompt=True,29 tokenize=True,30 return_dict=True,31 return_tensors="pt",32).to(model.device).to(torch.bfloat16)333435with torch.inference_mode():36 outputs = model.generate(**inputs, max_new_tokens=64)3738outputs = tokenizer.batch_decode(outputs)
Data used for model training and how the data was processed.
Training Dataset
These models were trained on a dataset of text data that includes a wide variety
of sources. The 27B model was trained with 14 trillion tokens, the 12B model was
trained with 12 trillion tokens, 4B model was trained with 4 trillion tokens and
1B with 2 trillion tokens. Here are the key components:
Web Documents: A diverse collection of web text ensures the model is
exposed to a broad range of linguistic styles, topics, and vocabulary. The
training dataset includes content in over 140 languages.
Code: Exposing the model to code helps it to learn the syntax and
patterns of programming languages, which improves its ability to generate
code and understand code-related questions.
Mathematics: Training on mathematical text helps the model learn logical
reasoning, symbolic representation, and to address mathematical queries.
Images: A wide range of images enables the model to perform image
analysis and visual data extraction tasks.
The combination of these diverse data sources is crucial for training a powerful
multimodal model that can handle a wide variety of different tasks and data
formats.
Data Preprocessing
Here are the key data cleaning and filtering methods applied to the training
data:
CSAM Filtering: Rigorous CSAM (Child Sexual Abuse Material) filtering
was applied at multiple stages in the data preparation process to ensure
the exclusion of harmful and illegal content.
Sensitive Data Filtering: As part of making Gemma pre-trained models
safe and reliable, automated techniques were used to filter out certain
personal information and other sensitive data from training sets.
Additional methods: Filtering based on content quality and safety in
line with our policies.
Implementation Information
Details about the model internals.
Hardware
Gemma was trained using Tensor Processing Unit (TPU) hardware (TPUv4p,
TPUv5p and TPUv5e). Training vision-language models (VLMS) requires significant
computational power. TPUs, designed specifically for matrix operations common in
machine learning, offer several advantages in this domain:
Performance: TPUs are specifically designed to handle the massive
computations involved in training VLMs. They can speed up training
considerably compared to CPUs.
Memory: TPUs often come with large amounts of high-bandwidth memory,
allowing for the handling of large models and batch sizes during training.
This can lead to better model quality.
Scalability: TPU Pods (large clusters of TPUs) provide a scalable
solution for handling the growing complexity of large foundation models.
You can distribute training across multiple TPU devices for faster and more
efficient processing.
Cost-effectiveness: In many scenarios, TPUs can provide a more
cost-effective solution for training large models compared to CPU-based
infrastructure, especially when considering the time and resources saved
due to faster training.
JAX allows researchers to take advantage of the latest generation of hardware,
including TPUs, for faster and more efficient training of large models. ML
Pathways is Google's latest effort to build artificially intelligent systems
capable of generalizing across multiple tasks. This is specially suitable for
foundation models, including large language models like these ones.
Together, JAX and ML Pathways are used as described in the
paper about the Gemini family of models; "the 'single
controller' programming model of Jax and Pathways allows a single Python
process to orchestrate the entire training run, dramatically simplifying the
development workflow."
Evaluation
Model evaluation metrics and results.
Benchmark Results
These models were evaluated against a large collection of different datasets and
metrics to cover different aspects of text generation:
Ethics and safety evaluation approach and results.
Evaluation Approach
Our evaluation methods include structured evaluations and internal red-teaming
testing of relevant content policies. Red-teaming was conducted by a number of
different teams, each with different goals and human evaluation metrics. These
models were evaluated against a number of different categories relevant to
ethics and safety, including:
Child Safety: Evaluation of text-to-text and image to text prompts
covering child safety policies, including child sexual abuse and
exploitation.
Content Safety: Evaluation of text-to-text and image to text prompts
covering safety policies including, harassment, violence and gore, and hate
speech.
Representational Harms: Evaluation of text-to-text and image to text
prompts covering safety policies including bias, stereotyping, and harmful
associations or inaccuracies.
In addition to development level evaluations, we conduct "assurance
evaluations" which are our 'arms-length' internal evaluations for responsibility
governance decision making. They are conducted separately from the model
development team, to inform decision making about release. High level findings
are fed back to the model team, but prompt sets are held-out to prevent
overfitting and preserve the results' ability to inform decision making.
Assurance evaluation results are reported to our Responsibility & Safety Council
as part of release review.
Evaluation Results
For all areas of safety testing, we saw major improvements in the categories of
child safety, content safety, and representational harms relative to previous
Gemma models. All testing was conducted without safety filters to evaluate the
model capabilities and behaviors. For both text-to-text and image-to-text, and
across all model sizes, the model produced minimal policy violations, and showed
significant improvements over previous Gemma models' performance with respect
to ungrounded inferences. A limitation of our evaluations was they included only
English language prompts.
Usage and Limitations
These models have certain limitations that users should be aware of.
Intended Usage
Open vision-language models (VLMs) models have a wide range of applications
across various industries and domains. The following list of potential uses is
not comprehensive. The purpose of this list is to provide contextual information
about the possible use-cases that the model creators considered as part of model
training and development.
Content Creation and Communication
Text Generation: These models can be used to generate creative text
formats such as poems, scripts, code, marketing copy, and email drafts.
Chatbots and Conversational AI: Power conversational interfaces
for customer service, virtual assistants, or interactive applications.
Text Summarization: Generate concise summaries of a text corpus,
research papers, or reports.
Image Data Extraction: These models can be used to extract,
interpret, and summarize visual data for text communications.
Research and Education
Natural Language Processing (NLP) and VLM Research: These
models can serve as a foundation for researchers to experiment with VLM
and NLP techniques, develop algorithms, and contribute to the
advancement of the field.
Language Learning Tools: Support interactive language learning
experiences, aiding in grammar correction or providing writing practice.
Knowledge Exploration: Assist researchers in exploring large
bodies of text by generating summaries or answering questions about
specific topics.
Limitations
Training Data
The quality and diversity of the training data significantly
influence the model's capabilities. Biases or gaps in the training data
can lead to limitations in the model's responses.
The scope of the training dataset determines the subject areas
the model can handle effectively.
Context and Task Complexity
Models are better at tasks that can be framed with clear
prompts and instructions. Open-ended or highly complex tasks might be
challenging.
A model's performance can be influenced by the amount of context
provided (longer context generally leads to better outputs, up to a
certain point).
Language Ambiguity and Nuance
Natural language is inherently complex. Models might struggle
to grasp subtle nuances, sarcasm, or figurative language.
Factual Accuracy
Models generate responses based on information they learned
from their training datasets, but they are not knowledge bases. They
may generate incorrect or outdated factual statements.
Common Sense
Models rely on statistical patterns in language. They might
lack the ability to apply common sense reasoning in certain situations.
Ethical Considerations and Risks
The development of vision-language models (VLMs) raises several ethical
concerns. In creating an open model, we have carefully considered the following:
Bias and Fairness
VLMs trained on large-scale, real-world text and image data can
reflect socio-cultural biases embedded in the training material. These
models underwent careful scrutiny, input data pre-processing described
and posterior evaluations reported in this card.
Misinformation and Misuse
VLMs can be misused to generate text that is false, misleading,
or harmful.
This model card summarizes details on the models' architecture,
capabilities, limitations, and evaluation processes.
A responsibly developed open model offers the opportunity to
share innovation by making VLM technology accessible to developers and
researchers across the AI ecosystem.
Risks identified and mitigations:
Perpetuation of biases: It's encouraged to perform continuous
monitoring (using evaluation metrics, human review) and the exploration of
de-biasing techniques during model training, fine-tuning, and other use
cases.
Generation of harmful content: Mechanisms and guidelines for content
safety are essential. Developers are encouraged to exercise caution and
implement appropriate content safety safeguards based on their specific
product policies and application use cases.
Misuse for malicious purposes: Technical limitations and developer
and end-user education can help mitigate against malicious applications of
VLMs. Educational resources and reporting mechanisms for users to flag
misuse are provided. Prohibited uses of Gemma models are outlined in the
Gemma Prohibited Use Policy.
Privacy violations: Models were trained on data filtered for removal
of certain personal information and other sensitive data. Developers are
encouraged to adhere to privacy regulations with privacy-preserving
techniques.
Benefits
At the time of release, this family of models provides high-performance open
vision-language model implementations designed from the ground up for
responsible AI development compared to similarly sized models.
Using the benchmark evaluation metrics described in this document, these models
have shown to provide superior performance to other, comparably-sized open model
alternatives.
