Google has introduced Gemma 4, a new generation of open large language models that builds on the technological foundations of its earlier Gemini models.
The latest release brings significant advancements in reasoning capabilities and agent-driven workflows, extending cutting-edge AI innovations to open ecosystems.
The models also support training across more than 140 languages, enhancing their global usability.
Gemma 4 arrives in four variants, each designed for different computing environments. These include the Effective 2B (E2B), Effective 4B (E4B), 26B Mixture of Experts, and 31B Dense models.
The larger models focus on high-performance computing and advanced intelligence tasks, while the smaller variants cater to smartphones, edge devices, and Internet of Things systems.
Unlike chatbot-oriented systems, Gemma functions as an AI processing engine that developers can download and run locally. This approach offers greater flexibility for customisation and fine-tuning, while also improving data privacy and reducing reliance on cloud-based infrastructure.
Google has collaborated with major industry partners, including Qualcomm, MediaTek, and Nvidia, to optimise performance across platforms.
Released under the Apache 2.0 licence, the models allow developers and organisations to modify and deploy them with minimal restrictions.
According to Clement Farabet, the models deliver exceptional intelligence relative to their size, building on strong developer adoption worldwide.
Benchmark evaluations also place the 31B model among the top-performing open systems globally.
With capabilities spanning text, audio, and image processing, along with extended context windows and efficient performance on advanced GPUs, Gemma 4 marks a major step in making powerful AI tools more accessible.
Also Read: Can You Use ChatGPT While Driving? It’s Now On Apple CarPlay
To read more such news, download Bharat Express news apps
