Google Introduced Gemma: An Open Version Of The Gemini AI model

Google has released new Gemma artificial intelligence models that third-party developers can use to create their own models, the company said in a blog post. The model was developed by DeepMind and other Google teams.

The lack of complexity of Gemma 2B and 7B can be compensated by their higher speed and low deployment cost. Open neural networks, according to Google, “significantly outperform larger models on key metrics” and “can run directly on a developer’s laptop or PC.” They will appear on the Kaggle, Hugging Face, NVIDIA NeMo and Google Vertex AI platforms. Gemmas will be available under a commercial license regardless of the size of the organization, the number of users or the type of project, but they are prohibited from being used for certain tasks, such as weapons development.

Gemma comes with the Responsible Generative AI Toolkit, which provides guidance for building AI applications, as well as a debugging tool that lets you examine model behavior and fix problems. Currently, Gemma 2B and Gemma 7B are best suited for English language tasks.

You can try Gemma completely free, and the ability to run it on your own PC makes the model more attractive than many others.