Gemma
by Google DeepMind
Google's open-weight models for running and fine-tuning on phones, laptops or servers.
About
Gemma is a family of open models from Google DeepMind, built from the same research as Gemini, in sizes for mobile devices up to workstations. Gemma 4 is multimodal, and there are variants for embeddings, safety classification, translation and medical tasks.
Demo video
Found on deepmind.google’s homepage. The video belongs to its owner.
More pictures
Further down the homepage Further down the homepage The picture the site shows when it’s shared
Taken from deepmind.google. The pictures belong to the site’s owner and are shown to identify the site.
Useful features
- Open-weight models in several sizes
- Multimodal input in Gemma 4
- Variants for embeddings, safety and medicine
- Runs locally, on mobile or in the cloud
Good for
- On-device and offline AI features
- Fine-tuning a model on your own data
- Self-hosted chat and agent backends
Pricing
Free. Free to download and use; Gemma 4 is under Apache 2.0 and earlier versions under Google's Gemma terms. Hosted use on Google Cloud or other providers is billed by those services.
More in AI developer tools
-
#1
Platform for sharing and running open AI models, datasets and demo apps.
-
#2
Meta's open-weight language models that you can download, fine-tune and run yourself.
-
#3
Enterprise AI models for text generation, embeddings and search reranking via API.
-
#5
Jamba open models, an API, and a gateway that cuts token costs for AI agents.
