Gemma 4: Byte for byte, the most capable open models
5 hours ago
- #Edge Computing
- #AI Models
- #Open Source
- Gemma 4 introduces Google's most intelligent open models with advanced reasoning and agentic workflows.
- The models come in four sizes: Effective 2B, Effective 4B, 26B Mixture of Experts, and 31B Dense.
- Gemma 4 delivers high intelligence-per-parameter, outperforming larger models and enabling frontier capabilities on less hardware.
- Edge models (E2B/E4B) prioritize multimodal features, low latency, and on-device utility for mobile and IoT devices.
- Models support advanced reasoning, function-calling, JSON output, code generation, vision/audio processing, and 140+ languages.
- Released under an Apache 2.0 license for flexibility, with tools for experimentation, customization, and deployment.
- Built on rigorous safety protocols, offering trusted, transparent AI for enterprises and developers.