Gemma

The Gemma repository describes an open-weight model family based on Gemini research and a JAX library for using and fine-tuning Gemma. It states that it is not an official Google product.

Features described on the official site

• The README distinguishes the model family from the gemma Python/JAX implementation.
• Describes multi-turn/multimodal sampling and fine-tuning, linking examples and model reports.
• The current example uses Gemma 4, with multiple versions described and separate CPU/GPU/TPU conditions.

Suggested starting workflow

• Define a model version, task and implementation library, checking weight terms first.
• Plan an authorized short sample and review memory, dependencies and inference configuration.
• Record model/library versions, outputs and evaluation methods separately before calculating costs.

These are editorial starting suggestions based on product purpose, not workflows tested by this site.

Before you use it

• Not tested hands-on here; descriptions come from official pages or project material.
• Pricing, regions, languages, commercial use, watermarks and export conditions have not been verified for a single plan.
• Developer documentation timed out, so the google-deepmind project README is used. Its not-an-official-Google-product statement is retained without inferring support or guarantees. Open weights do not establish identical commercial terms; older hardware examples are not applied to Gemma 4. No installation, download, fine-tuning or inference occurred.