Google Launches Gemma 4: Most Powerful Open-Source AI Model Family Yet Runs on Everything From Phones to IoT Devices
Summary
Google launches Gemma 4, its most powerful open-source AI model family yet, offering four sizes that run on everything from Android phones to IoT devices, with top models ranking in the global top 10 for open AI, supporting 140 languages, and available free under an Apache 2.0 license.
Key Points
- Google launches Gemma 4, its most capable open-source model family yet, available in four sizes — Effective 2B, Effective 4B, 26B MoE, and 31B Dense — designed to run on devices ranging from Android phones to IoT hardware.
- Gemma 4 delivers major improvements in reasoning, code generation, agentic workflows, multimodal processing, and supports 140 languages, with its 31B Dense model ranking #3 and its 26B MoE ranking #6 among open models on the Arena AI leaderboard.
- Released under an Apache 2.0 license and available on Hugging Face, Kaggle, and Ollama, Gemma 4 builds on the original Gemma family's 400 million downloads since 2024, reinforcing Google's strategy of keeping open models available to developers while maintaining proprietary control over its flagship Gemini lineup.