Gemini Nano 4
An on-device large model introduced by Google that runs locally on mobile devices, supporting offline inference, zero cloud cost, and on-device data privacy protection.
Timeline (last 90 days)
In a tech talk, Google engineers Amy and Alice introduced building audio-first AI experiences for smart glasses using Gemini Nano 4 and ML Kit, and provided an open-source sample, XR Language Teacher.
Gemini Nano 4 is built on the open-source model Gemma 4 and is invoked directly on supported flagship Android devices through the ML Kit API; the model is delivered by the Android system, so apps do not need to bundle the model.
Gemini Nano 4's training covers more than 140 languages, can handle idioms and formal and informal registers, and follows system prompts to a high degree.
Gemini Nano 4 is a Gemma model built and run through AI Core; AI Core is a system-level inference service for on-device AI on the Android platform, where models are shared across apps and automatically updated with the system.
While maintaining low latency and offline capability, Gemini Nano 4 improves multimodal understanding and reasoning quality, but it remains significantly weaker than cloud-based large models and is suitable for tasks such as summarization, classification, image recognition, and simple conversation.
All Facts (5)
Gemini Nano 4 is built on the open-source model Gemma 4 and is invoked directly on supported flagship Android devices through the ML Kit API; the model is delivered by the Android system, so apps do not need to bundle the model.
50%UnverifiedGemini Nano 4's training covers more than 140 languages, can handle idioms and formal and informal registers, and follows system prompts to a high degree.
50%UnverifiedGemini Nano 4 is a Gemma model built and run through AI Core; AI Core is a system-level inference service for on-device AI on the Android platform, where models are shared across apps and automatically updated with the system.
50%UnverifiedWhile maintaining low latency and offline capability, Gemini Nano 4 improves multimodal understanding and reasoning quality, but it remains significantly weaker than cloud-based large models and is suitable for tasks such as summarization, classification, image recognition, and simple conversation.
50%UnverifiedIn a tech talk, Google engineers Amy and Alice introduced building audio-first AI experiences for smart glasses using Gemini Nano 4 and ML Kit, and provided an open-source sample, XR Language Teacher.
50%