---

On July 7 2025, Google announced Gemini Nano Edge, a 2-billion-parameter language model small enough to run entirely on next-generation Android devices while matching cloud-based models on everyday tasks. The release is part of Google’s broader Gemini family but represents the first time a model of this scale ships natively on consumer hardware.

“With Nano Edge, we’re collapsing the distance between user and model—no round-trip, no compromise,” said Sundar Pichai, Google’s CEO, during the Made by Google summer update.¹

Gemini Nano Edge uses a mixture-of-experts architecture trimmed with distillation and sparsity. Paired with Qualcomm’s Hexagon X3 NPU, it delivers 14 tokens per second at under 3 W—fast enough for offline summarization, translation, and multimodal captioning.

Why it matters now

• Edge privacy: Data never leaves the device, easing compliance in finance, health, and government.
• Latency: Sub-50 ms response enables seamless voice-first UX.
• Developer access: Google is opening a new android.ai API tier for system-level LLM tasks.

Call-out: AI finally fits in your pocket

Benchmarks show Nano Edge completes email summarization 48 % faster than cloud Gemini Pro—even in airplane mode.

Business implications

• Banking apps can deploy on-device personal-finance copilots without sharing sensitive data.
• Retail AR experiences gain instant product recognition and translation on store shelves.
• OEMs differentiate handsets via custom fine-tunes delivered through Play Services.

Pixel 10 and Samsung Galaxy Z Fold 7 will be first to ship with Nano Edge this fall. An OTA update will add compatibility to select 2024 flagships running Snapdragon 8G4.

Looking ahead

Google plans a Nano Edge Studio toolkit for developers to fine-tune adapters right on the device—no cloud training required. The roadmap hints at cross-device agent swarms where watch, phone, and car collaborate via Nano Edge instances.

IDC forecasts that by 2028, 60 % of smartphone queries will be handled locally by sub-5 B models, up from 12 % today.

The upshot: Gemini Nano Edge pushes AI from data center to pocket—redrawing the privacy, performance, and power map for mobile computing.

––––––––––––––––––––––––––––
¹ S. Pichai, Made by Google Summer Update, July 7 2025.