Gemma 3 1B (local general‑purpose model)

A high-efficiency 1-billion parameter multimodal model built for low-latency local execution on edge devices.

Gemma 3 1B brings native multimodal reasoning (processing both text and images) to local hardware without requiring cloud access. It features a 128k-token context window and runs smoothly on consumer-grade CPUs or mobile chipsets (like the Snapdragon 8 Gen 3). This model excels at privacy-focused tasks: summarizing local documents, powering offline chat interfaces, and performing basic visual analysis. By prioritizing a small memory footprint, it delivers sub-second response times for developers building lightweight, secure applications.

https://ai.google.dev/gemma