Google launches open-source model Gemma 4: How to try it
Google just released the latest version of its open AI model, Gemma 4, on Thursday. Above all, Gemma 4 is a completely open source model under the Apache 2.0 license, which is generally not the case for 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 the foundation for complete developer flexibility and digital sovereignty; giving you complete control over your data, infrastructure, and models,” reads the Google blog. “It lets you build freely and deploy securely in any environment, whether on-premises or in the cloud. »
Most people have probably heard of Google’s popular Gemini AI model, thanks to the ubiquitous AI chatbot that has been integrated into many Google products.
Gemma is also a Large Language Model (LLM) and was developed using the same technology and research that Google DeepMind used to create Gemini 3.
Google calls Gemma 4 its “highest performing” open AI model to date.
Gemma versus Gemini?
So how is Gemma different from Gemini?
Gemini is Google’s proprietary subscription AI product and the name of Google’s family of multi-modal AI models. Gemini has been integrated into virtually all of Google’s core products, including Google Search, Gmail, Google Docs, and Google Cloud.
Gemma 4, however, is an open AI model, meaning the code and data it is trained on is shared with its user base. Gemma AI models can be run from a user’s local hardware, even without an internet connection. Anyone can download Gemma 4 and run it on their device for free. These open AI models provide a more private and secure experience because none of the chats, downloaded files, or responses are shared with any third party.
Crushable speed of light
Developers could use open AI models like Gemma 4 to integrate AI into their own applications without the need for recurring subscription fees.
What is Gemma 4?
Gemma 4 brings advanced capabilities to Google’s family of open AI models.
According to Google’s announcement, Gemma 4 is now capable of advanced reasoning, which includes multi-step planning and deep logic. Google says it has made “significant improvements to math and instruction-following tests that require it” with Gemma 4.
Gemma 4 now also supports the processes required for agent workflows and localizes AI coding assistance. Additionally, Gemma 4 can process audio and video for speech recognition and interpretation of visuals such as graphics.
Gemma 4 is available in four sizes depending on the number of weights used to power the model: two billion, four billion, 26 billion and 31 billion.
Hugging Face reports that these open-weight models are available in pre-trained and instruction-friendly variants, providing even more flexibility for developers.
The AI model was trained on more than 140 languages and has a popup of up to 256,000 tokens, according to Google. (The smaller E2B and E4B variants have a pop-up of 128,000, however.)
Gemma 4 is now open And open source
Now, open doesn’t mean open source when it comes to AI models.
Previous iterations of Gemma were open (meaning the training datasets are publicly available) but were still bound by Google’s terms, even though users were allowed to download the model to their device. Although users could modify the local LLM, they still had to follow Google’s policies regarding its use and redistribution.
With Gemma 4, Google has now made the model open And open source.
Google distributes Gemma 4 under the popular Apache 2.0 open source software license.
Under this license, anyone can download and modify Gemma 4 and use it for any purpose, whether personal or commercial. Gemma 4 can also be redistributed without any royalty requirements. Basically, the only requirement of the Apache 2.0 license is attribution, and the license must be distributed alongside the AI model.
Looking for Gemma 4? Hot to try it.
Gemma 4 can be found in Google AI Studio and can be downloaded from third parties such as Hugging Face, Kaggle or Ollama.
