Gemma 4 Debuts with 140-Language Support
News Desk
San Francisco: Google is pushing deeper into the open AI ecosystem with the launch of its latest Gemma 4 models, aiming to empower developers with highly efficient, multilingual tools capable of running from cloud systems to mobile devices.
Developed by Google DeepMind, the new release introduces a range of model sizes—including effective 2B (E2B), effective 4B (E4B), a 26B Mixture of Experts (MoE), and a 31B dense model—designed to support advanced reasoning and agent-based workflows under the permissive Apache 2.0 license.
The company says Gemma 4 represents its most capable open model to date, with the 31B variant currently ranked among the top-performing open models globally on the Arena AI text leaderboard. The 26B model has also demonstrated strong performance, reportedly outperforming systems significantly larger in scale.
A key focus of the update is accessibility and versatility. The models support over 140 languages and offer extended context windows of up to 256,000 tokens, enabling developers to process large datasets such as full codebases or lengthy documents in a single interaction.
Read More: https://thepenpk.com/nano-qr-code-could-store-terabytes-for-centuries/
Gemma 4 also introduces enhanced capabilities for building autonomous AI systems, including native function calling, structured JSON outputs, and system command integration. These features allow developers to create agents that can interact with external tools and APIs more seamlessly.
In addition to text, the models support multimodal inputs—including images and video—while smaller variants (E2B and E4B) add native audio processing for speech recognition.
Notably, Google has optimised the lighter models for edge computing, enabling them to run offline on devices such as smartphones, Raspberry Pi, and NVIDIA Jetson platforms with minimal latency. The company collaborated with hardware partners like Qualcomm and MediaTek to improve mobile performance.
Since the initial launch of the Gemma family, the models have been downloaded more than 400 million times, spawning a growing ecosystem of over 100,000 community-built variants.
Gemma 4 is now available through platforms including Hugging Face, Kaggle, Ollama, and Google AI Studio, reflecting Google’s broader push to expand access to advanced AI tools for developers worldwide.
