Google launches open-source model Gemma 4: How to try it
Mashable • • general
Key Points:
- Google has released Gemma 4, its latest open-source AI model licensed under Apache 2.0, allowing full developer flexibility and local usage on billions of Android devices and some laptop GPUs.
- Unlike Google's proprietary Gemini AI models integrated into many Google products, Gemma 4 is fully open-source, enabling users to run it locally without internet and without sharing data with third parties, enhancing privacy and security.
- Gemma 4 offers advanced capabilities such as multi-step reasoning, deep logic, improved math and instruction-following, audio and video processing, and supports agentic workflows and AI coding assistance.
- The model is available in four sizes ranging from 2 billion to 31 billion parameters, supports over 140 languages, and has a large context window up to 256,000 tokens, with both pre-trained and instruction-tuned variants.
- Gemma 4 can be accessed via Google AI Studio or downloaded from platforms like Hugging Face, Kaggle, and Ollama, and its Apache 2.0 license permits free use, modification, and redistribution with attribution.
