
Google launches Gemma 4 open AI model, it can run on your smartphone
Google has launched its latest open-weight AI model called Gemma 4. According to the company, Gemma 4 is the best open model in its size out there. And it can even run locally on your Android smartphone. Here is everything you need to know.

Google has launched its latest open AI model, called Gemma 4. Unlike most frontier models like Gemini or OpenAI’s GPT 5.4, Gemma 4 is an open-source model that can run locally on a wide range of devices, including your Android smartphone. According to the company, Gemma 4 gives developers the freedom of open models with the same level of functionality as proprietary models.
Google DeepMind CEO Demis Hassabis announced the release on X. He claimed that the 4 sizes of Gemma 4 were “the best open models in the world for their respective sizes.” Here are all the details.
Unlike Gemini, these models are open-source. That is, anyone can download Gemma 4 and run it locally on their device for free. This also gives you more privacy as the locally-run model will not share any data with Google or any other third party.
White House policy advisor Sriram Krishnan stated that open-source models were crucial for the West. He wrote, “Open source models are a key front for the west to have a lead on and this is a very key addition to the effort.”
What is Gemma 4?
Gemma 4 consists of 4 sizes of purpose-built models for advanced reasoning and agentic workflows. The Gemma 4 family includes four models: Effective 2 Billion (E2B) and Effective 4 Billion (E4B) designed for mobile and edge devices; 26 Billion Mixture of Experts (MoE) focusing on low latency; and 31 Billion Dense, which offers top raw performance.
The 31B model ranks third globally among open models on the Arena AI leaderboard, outperforming models 20 times its size.
According to Google, Gemma 4 brings a big leap in advanced capabilities such as multi-step planning and complex logic. The Gemma 4 models also support agentic workflows that can allow you to delegate your tasks to an AI agent. You can build a team of AI agents with Gemma 4 that can then do your tasks, think something like OpenClaw.
Gemma 4 on Android
Gemma 4 is designed to run on a range of hardware, going from large computers with high performance GPUs to even regular Android smartphones. According to Google, Gemini 4 is the base model for the next-generation of Gemini Nano – the foundation AI model on Android.
This could allow billions of smartphones to run the AI model locally, without needing an internet connection. Usually, AI models like ChatGPT or Gemini run via the cloud, and you cannot access these models on your phone without an internet connection.
Google goes open-source with Gemma 4
According to Google, since the first generation’s launch, developers have downloaded Gemma over 400 million times, creating a community ecosystem with more than 100,000 variants. The models are released under an Apache 2.0 licence, which gives developers the complete flexibility to use, modify and deploy the models.
Do note that earlier Gemma models were not open-source, but open-weight, that is, only their training datasets are publicly available. And users still had to run those models as per Google’s terms.
What else can Gemma 4 do?
Gemma 4 also provides high-quality offline code generation, allowing you to have a local AI coding assistant. All models can process images and video, while the smaller E2B and E4B also handle audio input for speech recognition.
Gemma 4 supports longer context windows than previous generations. With edge models processing up to 128,000 tokens and larger models up to 256,000 tokens, enabling handling of long documents or code repositories in a single prompt. The models are trained in over 140 languages.
Gemma 4 is accessible through Google AI Studio and can be downloaded from platforms such as Hugging Face, Kaggle and Ollama.

