Google AI efforts are synonymous to Gemini. It is now an integral part of its most popular software and hardware products, including Worksuite. The company has released multiple open-source AI model under the Gemma brand for over a decade now.
Google released its Gemma AI model today. The company revealed its third-generation open-source AI models, which boasted some impressive claims. The Gemma 3 models are available in four variations — 1 billion, 2 billion, 4 billion and 12 billion parameters — designed to run on devices from smartphones to powerful workstations.
Ready for mobile devices
) Google claims that Gemma 3 is a world-leading single-accelerator algorithm, meaning it can be run on a single GPU/TPU unit instead of a cluster. Theoretically speaking, this means that a Gemma 3 model could run natively on the Pixel’s Tensor Processing Core unit (TPU), just as it does the Gemini Nano model on phones.
Gemma 3 has a major advantage over the Gemini AI models. It’s open source, so developers can package it and ship it to meet their specific needs in mobile apps and desktop applications. Gemma 3 supports 140 languages and 35 of them are included in a pre-trained package.
Gemma 3 can also understand text, images and videos, just like the Gemini 2.0 models. It is multi-multimdal. Gemma 3’s performance is said to be superior to other popular open-source AI systems such as DeepSeek V3, OpenAI o3 mini, and Meta Llama-405B.
Versatile and ready to deploy.
Gemma 3 provides a context window with 128,000 tokens. This is enough tokens to cover the cost of a 200-page book when pushed in as an input. Comparatively, the context window of Google’s Gemini Flash Lite 2.0 model is one million tokens. In the context AI models, a typical English language word is approximately equivalent to 1.3 tokens.
Gemma 3 also supports function calling and structured output, which essentially means it can interact with external datasets and perform tasks like an automated agent. The nearest analogy would be Gemini, and how it can get work done across different platforms such as Gmail or Docs seamlessly.
The latest open-source AI models from Google can either be deployed locally, or through the company’s cloud-based platforms such as the Vertex AI suite. Gemma 3 AI models are now available via the Google AI Studio, as well as third-party repositories such as Hugging Face, Ollama, and Kaggle.
Gemma 3 is part of an industry trend where companies are working on Large Language Models (Gemini, in Google’s case) and simultaneously pushing out small language models (SLMs), as well. Microsoft also follows a similar strategy with its open-source Phi series of small language models.
Small language models such as Gemma and Phi are extremely resource efficient, which makes them an ideal choice for running on devices such as smartphones. Moroever, as they offer a lower latency, they are particularly well-suited for mobile applications.
