Google has launched the Gemini Nano, the most streamlined version of its new Gemini AI family, directly on the Pixel 8 Pro smartphone. Gemini Nano represents Google’s first ever on-device AI model, allowing the phone itself to handle advanced capabilities like summarization, language understanding, and more without an internet connection.

Specifically, Gemini Nano powers two new Pixel 8 highlight features: “Summarize” in the Recorder transcription tool and suggested replies in Gboard. By keeping the AI computation restricted to the device, Google can enable these features to work offline while ensuring better latency and reliability than cloud-dependent solutions.

Gemini Nano is specially designed by Google to run efficiently on mobile chipsets compared to larger AI models. It utilizes the Tensor Processing Unit (TPU) in Google’s own Tensor G3 mobile processor featured in the Pixel 8. Additionally, Gemini Nano will also be compatible with AI accelerators from leading mobile silicon providers like Qualcomm, Samsung, and MediaTek.

With Gemini Nano demonstrating the possibilities of powerful AI on smartphones, Google is poised to unlock more advanced capabilities by leveraging future iterations of its efficient models. Areas like the Google Assistant, image processing, predictive typing, audio transcription and more could be enhanced directly on users’ phones for speed, privacy, and reliability. As software taps deeper into on-device AI, Pixel phones will blaze the trail.