Google previews Gemini Nano 4 for Android, arriving on flagship devices this year
Google has previewed Gemini Nano 4, a new on-device language model for Android, available now in early access via AICore Developer Preview. The model comes in two versions: Gemini Nano 4 Fast (3x faster than previous models, 60% less battery) and Gemini Nano 4 Full (higher reasoning capability). The models will launch on new flagship Android devices later this year.
Google Previews Gemini Nano 4 for Android, Launching on Flagship Devices This Year
Google has unveiled Gemini Nano 4, a new generation of on-device language models for Android, built on the foundation of Gemma 4. The model is available now in early access through the AICore Developer Preview and will ship on new flagship Android devices later in 2026.
Two Model Variants
Google is releasing two versions of Gemini Nano 4:
Gemini Nano 4 Fast (E2B variant): Optimized for speed, claimed to be 3x faster than the E4B model with lower latency and fastest response times.
Gemini Nano 4 Full (E4B variant): Designed for higher reasoning power and complex task handling, positioning it as the more capable version.
Performance Improvements
Compared to previous versions, Gemina 4 claims to deliver:
- 4x faster execution speed on Android
- 60% lower battery consumption
- Multimodal understanding across text, image, and audio
- Support for 140+ languages natively
Expanded Capabilities
Google highlights four key capability improvements in the new model:
Reasoning: Chain-of-thought and conditional logic now return higher-quality results. Example use case includes content moderation decisions with structured reasoning.
Math: Improved numerical reasoning for financial calculations, budgeting, and quantitative problem-solving.
Time understanding: Enhanced temporal reasoning for calendar, reminder, and alarm applications.
Image understanding: Better optical character recognition (OCR) for chart understanding, visual data extraction, and handwriting recognition.
Developer Access and Roadmap
Developers can sign up for the AICore Developer Preview immediately. Code written for Gemma 4 will automatically work on Gemini Nano 4-enabled devices, simplifying the development transition.
Google plans several updates during the preview period, including:
- Tool calling support
- Structured output
- System prompts
- Thinking mode in Prompt API
The company has not disclosed specific pricing, parameter counts, or detailed benchmark comparisons against competing on-device models.
What This Means
Gemini Nano 4 positions Google to compete more effectively in the on-device AI market, particularly against Apple's integration of language models into iOS. The 3x speed improvement and 60% battery reduction are material gains for Android manufacturers building AI-native devices. The multimodal capabilities and 140+ language support expand use cases beyond simple text completion. The spring 2026 availability window suggests Gemini Nano 4 will be integrated into Google's own Pixel flagships and Samsung's Galaxy S series by summer. Developers should note that early API design (thinking mode, structured outputs) is still in flux, making this preview period critical for feedback before general availability.
Related Articles
Google releases Gemma 4 family under Apache 2.0 license with 2B to 31B models
Google has released Gemma 4, a family of four open models ranging from 2B to 31B parameters, now available under the Apache 2.0 license for the first time. The 31B dense model ranks 3rd on the Arena AI Text Leaderboard, while the 26B mixture-of-experts variant ranks 6th, both outperforming significantly larger competitors. All models support multimodal inputs and are available on Hugging Face, Kaggle, and Ollama.
NVIDIA Optimizes Google Gemma 4 for Local Agentic AI on RTX and Spark
NVIDIA has optimized Google's Gemma 4 models for local deployment on RTX and Spark platforms, targeting the emerging wave of on-device agentic AI. The optimization enables small, efficient models to access real-time local context for autonomous decision-making without cloud dependency.
Google DeepMind releases Gemma 4 with 4 model sizes, 256K context, and multimodal reasoning
Google DeepMind released Gemma 4, a family of open-weights multimodal models in four sizes: E2B (2.3B effective), E4B (4.5B effective), 26B A4B (3.8B active), and 31B (30.7B parameters). All models support text and image input with 128K-256K context windows, while E2B and E4B add native audio capabilities and reasoning modes across 140+ languages.
Google DeepMind releases Gemma 4 open models with multimodal capabilities and 256K context window
Google DeepMind released the Gemma 4 family of open-source models with multimodal capabilities (text, image, audio, video) and context windows up to 256K tokens. Four distinct model sizes—E2B (2.3B effective parameters), E4B (4.5B effective), 26B A4B (3.8B active), and 31B—are available under the Apache 2.0 license, with instruction-tuned and pre-trained variants.
Comments
Loading...