Geek Out Time: LoRA Fine-Tuning new Gemma-3–270M on a Free Colab GPU

The Gemma family is Google’s initiative to provide lightweight, open-weight models for developers and researchers. The name “Gemma” comes from the Latin for precious stone , and these models are built from the same research foundation as Google’s closed-source Gemini models.

The philosophy is to enable developers to build and deploy generative AI on a wide range of hardware — from cloud GPUs to laptops and even on-device mobile phones. This is a direct response to the community’s demand for more accessible and efficient AI models.

The Release of Gemma 3 270M

The Gemma-3–270M is the smallest in the Gemma 3 lineup, released alongside larger models that scale up to 27B parameters and multimodal capabilities.

This is an excerpt — the full article continues on Medium.