Gemma 3 at a glance
Gemma 3 came in 1B, 4B, 12B and 27B sizes with a 128K context window, function calling, support for over 140 languages and vision in the larger models.
What changed with Gemma 3
On March 12, 2025, Google released Gemma 3, a family of lightweight open models built from the research behind Gemini 2.0, in 1B, 4B, 12B and 27B sizes. The 4B, 12B and 27B models accept images, all offer a 128K-token context window, and Google shipped official quantized versions to cut memory needs.
Google said the 27B model beat much larger open models in early LMArena preference tests while fitting on a single GPU or TPU host. Gemma 3 supports function calling and structured output, works out of the box in over 35 languages with pretrained support for more than 140, and is available through Hugging Face, Kaggle, Ollama, Google AI Studio and Vertex AI. ShieldGemma 2, a 4B image-safety classifier, launched alongside it.
Gemma 3 in numbers
Values are copied from the linked source on the date shown. They are not IA Compa measurements.
- Context windowas of · Source for Context window: blog.google (opens in a new tab)
- 128K tokens
How it compares
Other models in the same category that we have covered, using each model's latest recorded value. (Models)
Where to compare Gemma 3
Independent leaderboards and official resources for comparing this model yourself. Rankings change often; each site shows its own current data.
General leaderboards
Why Gemma 3 matters for businesses
For businesses, Gemma 3 made it realistic to run a capable multilingual model on one workstation or private server, which helps when client data cannot leave your own infrastructure.
Sources for this story
- Gemma 3: Google's new open model based on Gemini 2.0 (opens in a new tab)Official
Google · blog.google
- google/gemma-3-27b-it (opens in a new tab)Official
Hugging Face · huggingface.co
- Gemma 3 27B model page (opens in a new tab)Benchmark
Artificial Analysis · artificialanalysis.ai
Last verified: . IA Compa summarizes public announcements for business readers. We are not affiliated with the companies mentioned. Always confirm details, pricing, and availability with the official source.
- #llm
- #gemma
- #open-weights
- #multilingual
