Gemma 3
A collection of lightweight, state-of-the-art open models built from the same research and technology that powers our Gemini 2.0 models
Handle complex tasks
Gemma 3's 128K-token context window lets your applications process and understand vast amounts of information enabling more sophisticated AI capabilities.
Go global, instantly
Communicate effortlessly across languages with Gemma 3's unparalleled multilingual capabilities. Develop applications that reach a global audience, with support for over 140 languages.
Understand words and images
Easily build applications that analyze images, text, and video opening up new possibilities for interactive and intelligent applications.
Build with the world's best single-accelerator model
Start building with Gemma
Advance research with Gemma
Discover a growing collection of specialized Gemma models for advanced research.
ShieldGemma 2 new
PaliGemma 2 New
Gemma Scope
Visualize Gemma 3 benchmarks
Interact with the chart to see Gemma 3's results across a range of LLM benchmarks.
MMLU-Pro
The MMLU benchmark is a test that measures the breadth of knowledge and problem-solving ability acquired by large language models during pretraining.
LiveCodeBench
Assesses code generation capabilities on real-world coding problems from platforms like LeetCode and Codeforces.
dev
Bird-SQL
Tests a model's ability to translate natural language questions into complex SQL queries across various domains.
GPQA Diamond
Challenges models with difficult questions written by Ph.D. holders in biology, physics, and chemistry.
SimpleQA
Evaluates a model's ability to answer simple, factual questions with short phrases.
FACTS Grounding
Evaluates if LLM responses are factually accurate and detailed enough, based on given input documents.
MATH
MATH evaluates a language model's ability to solve complex mathematical word problems, requiring reasoning, multi-step problem-solving, and the understanding of mathematical concepts.
HiddenMath
An internal holdout set of competition math problems.
val
MMMU
Evaluates multimodal understanding and reasoning across various disciplines requiring college-level knowledge.
100%
75%
50%
25%
0%
100%
75%
50%
25%
0%
*See the technical report for details on performance with other methodologies. Read the technical report
Explore the Gemmaverse
A vast ecosystem of community-created Gemma models and tools, ready to power and inspire your innovation
Choose your deployment target
Mobile
Deploy on-device with Google AI Edge
Deploy directly to devices for low-latency, offline functionality. Ideal for applications requiring real-time responsiveness and privacy, such as mobile apps, IoT devices, and embedded systems.
Web
Integrate seamlessly into web applications
Empower your websites and web services with advanced AI capabilities, enabling interactive features, personalized content, and intelligent automation.
Cloud
Scale effortlessly with cloud infrastructure
Leverage the scalability and flexibility of the cloud to handle large-scale deployments, demanding workloads, and complex AI applications.
Accelerating academic research with Google Cloud credits
Advance your research with Gemma 3 models in Google Cloud. Apply now to receive Google Cloud credits to push the boundaries of your research and contribute to the advancement of the scientific community.
Selected researchers will receive Google Cloud credits to accelerate their scientific endeavors
Apply now