Google AIApril 2, 20263 min read

Google Releases Gemma 4 Family for Local and Fine-Tuned Deployments

Google's open-weight Gemma 4 family (26B and 31B parameter models) is now available on AI Studio — designed for developers needing local or customized AI deployments.

Google released the Gemma 4 model family on April 2, 2026, extending its open-weight AI ecosystem for developers who need to run models locally or fine-tune them on custom datasets.

Models released:

gemma-4-26b-a4b-it: 26B parameter instruction-tuned model, optimized for efficiency

gemma-4-31b-it: 31B parameter model with stronger reasoning and instruction following

Android AICore preview: Gemma 4 will be available through the AICore Developer Preview for Android, enabling on-device AI without cloud API calls — critical for privacy-sensitive applications.

Fine-tuning: Both models support full fine-tuning and LoRA adapters, making them suitable for domain-specific applications like legal document analysis, medical coding, or custom CRM workflows. For teams that want Gemini-quality outputs without sending data to Google's servers, Gemma 4 is the strongest open-weight option available.

Source:android-developers.googleblog.com

Want automation like this for your business?

Book a free call and we'll show you exactly what's possible for your setup.

Book a free call →