Gemma
In plain wordsGoogle's family of free-to-download AI language models: smaller relatives of its Gemini systems, compact enough to run on a single computer and, in some versions, on a phone.
A family of open-weightopen-weight model, n.: An AI model whose trained internal numbers are published for anyone to download and run on their own computers, even if the data and code used to make it stay private. language modelslarge language model, n.: A computer program trained on enormous amounts of text to predict the words that come next. At vast scale, that simple skill lets it chat, answer questions, translate, summarize and write code. that Google, chiefly through Google DeepMindGoogle DeepMind, prop. n.: Google's main artificial intelligence laboratory, formed in 2023 by joining DeepMind and Google Brain. It builds the Gemini models and scientific AI systems such as AlphaFold., has released since February 2024 as smaller siblings of its proprietary GeminiGemini, prop. n.: Google's family of AI models and its chatbot assistant, launched in December 2023. Gemini models handle text, images, audio and video and power many Google products. models, sharing their research lineage but sized to run on a single accelerator, a workstation or a phone. Gemma 1 (February 2024) offered decoder-only transformerstransformer, n.: The design behind today's chatbots and much of modern AI. It handles text by letting every word look at every other word and decide which of them matter for understanding it. of 2 and 7 billion parameters; Gemma 2 (mid-2024) came in 2B, 9B and 27B sizes, training the two smaller models by knowledge distillationknowledge distillation, n.: Training a smaller or cheaper 'student' AI to copy the behavior of a larger 'teacher' AI, so the student keeps much of the teacher's skill while being faster to run. from a larger teacher and alternating local and global attention layers; Gemma 3 (March 2025), from 1B to 27B parameters, added image input, context windows of up to 128,000 tokens and support for over 140 languages; Gemma 3n (mid-2025) targeted phones; and Gemma 4 (April 2026) — two small on-device models (E2B and E4B), a 26B mixture-of-experts model and a 31B dense model — moved the family to the permissive Apache 2.0 license. Specialized variants include CodeGemma for code, PaliGemma for vision and language, ShieldGemma safety classifiers, MedGemma for medicine, EmbeddingGemma for text embeddingstext embedding, n.: A list of numbers that captures what a piece of text means, so that texts with similar meanings get similar lists. Computers use them to search, sort and group text by meaning. and the Gemma ScopeGemma Scope, prop. n.: A free collection of ready-made concept-finding tools released by Google DeepMind for its open Gemma language models, so outside researchers can study their insides without training the tools themselves. interpretability tools.
We fine-tuned the 4B Gemma model on our support tickets and serve it from a single GPU.
Origin
Announced by Google on 21 February 2024 as 'open models' built from the research behind Gemini; the name comes from the Latin gemma, 'precious stone'. Google promotes the tens of thousands of community fine-tunes and derivatives published on Hugging Face as the 'Gemmaverse'.
Usage note
Gemma 1 to 3 shipped under custom Gemma Terms of Use with a prohibited-use policy, which critics said fell short of open sourceopen-source AI, n.: AI released, ideally with its code and details of its training data and not just the finished model, so that anyone may use, inspect, alter, share and even rebuild it. How open is open enough is debated.; the switch to Apache 2.0 with Gemma 4 removed those use restrictions, though training data remain unpublished. Not to be confused with GeminiGemini, prop. n.: Google's family of AI models and its chatbot assistant, launched in December 2023. Gemini models handle text, images, audio and video and power many Google products., Google's proprietary family.
Related
- GeminiGemini, prop. n.: Google's family of AI models and its chatbot assistant, launched in December 2023. Gemini models handle text, images, audio and video and power many Google products.
- open-weight modelopen-weight model, n.: An AI model whose trained internal numbers are published for anyone to download and run on their own computers, even if the data and code used to make it stay private.
- Gemma ScopeGemma Scope, prop. n.: A free collection of ready-made concept-finding tools released by Google DeepMind for its open Gemma language models, so outside researchers can study their insides without training the tools themselves.
- Google DeepMindGoogle DeepMind, prop. n.: Google's main artificial intelligence laboratory, formed in 2023 by joining DeepMind and Google Brain. It builds the Gemini models and scientific AI systems such as AlphaFold.
- small language modelsmall language model, n.: A language model built to be compact enough to run cheaply and quickly, even on a laptop or phone, trading some knowledge and skill for speed, privacy and lower cost.
- on-device AIon-device AI, n.: AI that runs directly on your own phone, laptop, car or headset rather than on a company's distant servers, so it can work offline, respond instantly and keep your data on the device.
- knowledge distillationknowledge distillation, n.: Training a smaller or cheaper 'student' AI to copy the behavior of a larger 'teacher' AI, so the student keeps much of the teacher's skill while being faster to run.
- LlamaLlama, prop. n.: A family of AI language models from Meta, first released in 2023, whose trained settings can be downloaded and run by anyone, making them the foundation for thousands of other AI projects.
- GeminiGemini, prop. n.: Google's family of AI models and its chatbot assistant, launched in December 2023. Gemini models handle text, images, audio and video and power many Google products.
Sources
- Gemma Team, Google DeepMind (2024). Gemma: Open Models Based on Gemini Research and Technology. arXiv.
- Gemma Team, Google DeepMind (2025). Gemma 3 Technical Report. arXiv.
- Banks, J. & Warkentin, T. (2024). Gemma: Introducing new state-of-the-art open models. Google blog, 21 February 2024. blog.google/technology/developers/gemma-open-models