Google announced its latest artificial intelligence (AI) invention, Gemma, following the debut of Gemini Pro 1.5 Pro a few days earlier. Gemma is a family of lightweight and open AI models developed by Google DeepMind, and Google claims that it is based on the same research and technology as Gemini.

Sundar Pichai, CEO of Google and Alphabet, Gemma posted on X: “Introducing Gemma, a family of lightweight, cutting-edge open models for their class, designed using the same research and technology as the Gemini models.”

Google Announces Gemma

Google has published Gemma in two variants: Gemma 2B and 7B, including pre-trained and instruction-tuned versions. In a blog post, Google stated that they both share technological and infrastructure components with Gemini. Gemma was created to help developers and academics build responsible artificial intelligence.

Gemma supports technologies commonly used by Google Cloud engineers, such as Colab and Kaggle notebooks. It also supports frameworks like JAX, PyTorch, Keras 3.0, and Hugging Face Transformer. According to Google, Gemma can run on a laptop, a workstation, or the Google Cloud.? According to the business, these new open models enable developers to build with and tweak Vertex AI and run it on Google Kubernetes Engine (GKE).

In terms of capabilities, Google claims that Gemma outperforms substantially larger models on important benchmarks while complying with our stringent standards for safe and responsible outcomes.

Gemma may be used with Vertex AI to create generative AI apps for lightweight tasks like text generation, summarization, and Q&A. It also supports real-time generative AI applications that necessitate minimal latency. Google also worked with NVIDIA to optimize Gemma for NVIDIA GPUs.

According to Google, developers may begin working with Gemma models on Google Cloud in Vertex AI and GKE right away.