Google launches Gemma 4, a new open-source model: How to try it

Source: Mashable Tech

Google just released the latest version of its open AI model, Gemma 4, on Thursday. Crucially, Gemma 4 is a fully open‑source model licensed under Apache 2.0, which is typically not the case with frontier models.

Open models can be run locally on users’ devices, and Google says Gemma 4 can be run on “billions of Android devices” and some laptop GPUs.

“This open‑source license provides a foundation for complete developer flexibility and digital sovereignty; granting you complete control over your data, infrastructure, and models. It allows you to build freely and deploy securely across any environment, whether on‑premises or in the cloud.” – Google blog post

Most people have likely heard of Google’s popular Gemini AI model, thanks to the ubiquitous AI chatbot integrated into many of Google’s products. Gemma is also a large language model (LLM) and was developed from the same technology and research that Google DeepMind used to build Gemini 3. Google is calling Gemma 4 its “most capable” open AI model yet.

Gemma vs. Gemini?

Gemini is Google’s proprietary subscription AI product and the name of its family of multimodal AI models. Gemini has been integrated into virtually all of Google’s core products, including Search, Gmail, Docs, and Cloud.

Gemma 4, however, is an open AI model, meaning that the code and data it was trained on are shared with its user base. Gemma models can be run on a user’s local hardware, even without an internet connection. Anyone can download Gemma 4 and run it on their device for free, providing a more private and secure experience—none of the chats, uploaded files, or answers are shared with a third party.

Developers can use open AI models like Gemma 4 to integrate AI into their own applications without recurring subscription costs.

What is Gemma 4?

Gemma 4 brings several advanced capabilities to Google’s open AI model family:

• Advanced reasoning: multi‑step planning and deep logic. Google reports “significant improvements in math and instruction‑following benchmarks.”
• Agentic workflows: supports processes required for autonomous agents and local AI coding assistance.
• Multimodal input: can process audio and video for speech recognition and interpret visuals such as charts.
• Model sizes: available in four variants—2 B, 4 B, 26 B, and 31 B parameters.
• Language coverage: trained on more than 140 languages.
• Context window: up to 256 000 tokens (the smaller E2B and E4B variants have a 128 000‑token window).

Hugging Face reports that these open‑weight models are available in pre‑trained and instruction‑tuned variants, offering additional flexibility for developers.

Gemma 4 is now open and open source

Previous iterations of Gemma were open‑weight (training datasets publicly available) but remained bound by Google’s terms of use. Users could modify the local LLM but still had to comply with Google’s redistribution rules.

With Gemma 4, Google has made the model both open and open source, distributing it under the Apache 2.0 license. This permits anyone to download, modify, and use the model for personal or commercial purposes, and to redistribute it without royalty requirements—only attribution is required.

Looking for Gemma 4? How to try it

Gemma 4 can be accessed through:

• Google AI Studio – gemma‑4 page
• Hugging Face – Gemma 4 collection
• Kaggle – Gemma 4 models
• Ollama – Gemma 4 library

These platforms provide pre‑trained weights and instruction‑tuned variants ready for download and local deployment.