Gemma 3n

Gemma 3n (commonly read as “Gemma 3 nano”) is a compact variant designed to bring Gemma 3 quality to constrained hardware. It’s text-in/text-out, instruction-tuned, and optimized for low latency while retaining practical reasoning, multilingual capability, and code generation. Developers can use function/tool calling, deterministic formats (JSON/Markdown), and guardrails to plug 3n into RAG pipelines, workflow automations, and small agent loops.

In practice, 3n targets single-GPU, CPU, or NPU/edge deployments with quantization support, making it easy to embed in apps that need quick answers without a server round-trip. Typical uses include chat copilots, document QA and summarization, UI/screenshot helpers (when paired with a vision encoder), and lightweight coding assistants. If you need maximum speed and minimal memory with “good enough” quality for everyday tasks, 3n is the sweet spot within the Gemma 3 lineup.