Gemma 3

Gemma 3 is Google's family of open multimodal models, distilled from the same research behind Gemini. The lineup spans from a 270M model light enough for phones to a 27B model that holds its own against far larger systems, with vision support arriving at 4B and above.

Every size handles text and understands 140+ languages, and the larger ones add image input, with context windows up to 128K tokens. That makes Gemma 3 a flexible base for everything from on-device assistants to cloud deployments.