Get the latest updates and exclusive AI content with our daily and weekly newsletters.
Google has announced that their new lightweight model series, Gemma 2, will be available to researchers and developers via Vertex AI starting next month. Initially, this series only had a 27-billion parameter model, but now, it also includes a 9-billion parameter model.
Gemma 2, which was first introduced at Google I/O in May, follows Gemma’s previous 2-billion and 7-billion parameter models from February. This new generation of models is designed to run on Nvidia’s latest GPUs or a single TPU host in Vertex AI. It’s specifically aimed at developers looking to integrate AI into their applications or edge devices like smartphones, IoT devices, and personal computers.
The two versions of Gemma 2 reflect current trends in the AI industry, where smaller and more lightweight models are becoming capable of handling various user demands. With 9-billion and 27-billion parameter options, Google offers developers flexibility in deciding whether to deploy these models on-device or via the cloud. Since the models are open-sourced, they can be easily customized and integrated into a range of projects.
There is also potential for the existing Gemma variants—such as CodeGemma, RecurrentGemma, and PaliGemma—to benefit from these new models.
Moreover, Google plans to introduce a 2.6-billion parameter model soon, aiming to balance lightweight accessibility with robust performance.
Gemma 2 can be accessed in Google AI Studio. Developers can download the model weights from platforms like Kaggle and Hugging Face, and researchers can use it for free through Kaggle or a free tier for Colab notebooks.