Stay up to date with the latest product announcements, technical deep dives,
hardware partnerships, and developer stories from the LiteRT and Google AI Edge
teams.

## Featured announcements

[![ML Drift universal GPU compute framework for on-device AI inference](https://developers.google.com/static/edge/litert/images/landing/ml-drift-banner.jpg)](https://developers.googleblog.com/ml-drift-next-gen-gpu-aiml-inference-at-the-edge/) Google Developers Blog GPU Acceleration

### [ML Drift: Next-Gen GPU AI/ML Inference at the Edge](https://developers.googleblog.com/ml-drift-next-gen-gpu-aiml-inference-at-the-edge/)

Accelerate on-device ML and generative AI inference across OpenGL ES, OpenCL, Metal, and WebGPU with Google's universal GPU compute framework.
Read blog post → [![LiteRT: The Universal Framework for On-Device AI](https://developers.google.com/static/edge/litert/images/landing/universal-framework.png)](https://developers.googleblog.com/litert-the-universal-framework-for-on-device-ai/) Google Developers Blog Official Announcement

### [LiteRT: The Universal Framework for On-Device AI](https://developers.googleblog.com/litert-the-universal-framework-for-on-device-ai/)

Google's unified on-device ML framework, evolving from TensorFlow Lite to deliver high-performance execution across mobile, desktop, web, and IoT platforms.
Read blog post → [![Official announcement rebranding TensorFlow Lite to LiteRT](https://developers.google.com/static/edge/litert/images/landing/LiteRT_Blog1.jpg)](https://developers.googleblog.com/tensorflow-lite-is-now-litert/) Google Developers Blog Official Announcement

### [TensorFlow Lite is now LiteRT](https://developers.googleblog.com/tensorflow-lite-is-now-litert/)

Google introduces LiteRT as the evolution of TensorFlow Lite, expanding support across PyTorch, JAX, and Keras for unified edge deployment.
Read blog post → [![LiteRT: Maximum Performance, Simplified](https://developers.google.com/static/edge/images/edge-article-3.jpg)](https://developers.googleblog.com/litert-maximum-performance-simplified/) Google Developers Blog Runtime Deep Dive

### [LiteRT: Maximum Performance, Simplified](https://developers.googleblog.com/litert-maximum-performance-simplified/)

Discover how the `CompiledModel` API streamlines on-device ML with automated accelerator selection, zero-copy buffer handling, and asynchronous execution across NPU, GPU, and CPU.
Read blog post → [![On-Device GenAI in Chrome, Chromebook Plus, and Pixel Watch with LiteRT-LM](https://developers.google.com/static/edge/images/edge-article-1.jpg)](https://developers.googleblog.com/on-device-genai-in-chrome-chromebook-plus-and-pixel-watch-with-litert-lm/) Google Developers Blog Generative AI

### [On-Device GenAI in Chrome, Chromebook Plus, and Pixel Watch with LiteRT-LM](https://developers.googleblog.com/on-device-genai-in-chrome-chromebook-plus-and-pixel-watch-with-litert-lm/)

Learn how LiteRT-LM brings small language models (SLMs) and on-device text generation directly into Chrome, Chromebook Plus devices, and Wear OS smartwatches.
Read blog post →

## Silicon \& hardware acceleration

Discover how LiteRT collaborates with silicon leaders to unlock peak hardware
efficiency and sub-millisecond inference latencies.

| Hardware Partner | Target Platform | Article Title | Link |
|---|---|---|---|
| **Qualcomm** | Snapdragon NPU | *Unlocking Peak Performance on Qualcomm NPU with LiteRT* | [Read Article →](https://developers.googleblog.com/unlocking-peak-performance-on-qualcomm-npu-with-litert/) |
| **MediaTek** | Dimensity NPU | *MediaTek NPU and LiteRT: Powering the Next Generation of On-Device AI* | [Read Article →](https://developers.googleblog.com/mediatek-npu-and-litert-powering-the-next-generation-of-on-device-ai/) |
| **Arm** | Mali \& Immortalis GPU | *How Arm and Google are Improving ML Drift Performance on Arm GPUs* | [Read Article →](https://developer.arm.com/community/arm-community-blogs/b/ai-blog/posts/how-arm-and-google-are-improving-ml-drift-performance-on-arm-gpus) |
| **Arm** | SME2 \& XNNPACK | *Accelerating On-Device AI: A Look at Arm and Google AI Edge Optimization* | [Read Article →](https://developers.googleblog.com/accelerating-on-device-ai-a-look-at-arm-and-google-ai-edge-optimization/) |
| **Intel** | Core Ultra NPU \& OpenVINO | *EmbeddingGemma v2 Day-0 on Intel NPU through LiteRT-LM and the OpenVINO Backend* | [Read Article →](https://medium.com/openvino-toolkit/embedding-gemma-v2-day-0-on-intel-npu-through-litert-lm-and-the-openvino-backend-2893009028b9) |
| **Intel** | Core Ultra NPU | *LiteRT Unlocks Core Ultra NPU Performance for AI PC* | [Read Article →](https://www.intel.com/content/www/us/en/developer/articles/community/litert-unlocks-core-ultra-npu-performance-for-aipc.html) |
| **Raspberry Pi** | Pi 5 CPU \& VideoCore GPU | *Mastering Edge AI on Raspberry Pi with LiteRT and Gemma* | [Read Article →](https://developers.googleblog.com/mastering-edge-ai-on-raspberry-pi-with-litert-and-gemma/) |
| **Google** | GPU (ML Drift) | *ML Drift: Next-Gen GPU AI/ML Inference at the Edge* | [Read Article →](https://developers.googleblog.com/ml-drift-next-gen-gpu-aiml-inference-at-the-edge/) |
| **Google** | Ecosystem Overview | *Building Real-World On-Device AI with LiteRT and NPU* | [Read Article →](https://developers.googleblog.com/building-real-world-on-device-ai-with-litert-and-npu/) |
| **Google Pixel** | Tensor TPU | *Google Tensor SDK Beta with LiteRT* | [Read Article →](https://developers.googleblog.com/google-tensor-sdk-beta-with-litert/) |

## Generative AI \& edge LLMs

[![Bring multimodal semantic search to the edge with EmbeddingGemma 2](https://developers.google.com/static/edge/litert/images/landing/embeddinggemma-2-banner.jpg)](https://developers.googleblog.com/google-ai-edge-with-embeddinggemma-2/) Google Developers Blog Multimodal Embeddings

### [Bring Multimodal Semantic Search to the Edge with EmbeddingGemma 2](https://developers.googleblog.com/google-ai-edge-with-embeddinggemma-2/)

Run private, low-latency multimodal retrieval and semantic search across text, vision, and audio on-device using EmbeddingGemma 2, MediaPipe, and LiteRT.
Read blog post → [![Antigravity SDK and LiteRT logos](https://developers.google.com/static/edge/litert/images/landing/antigravity-sdk-litert.png)](https://developers.googleblog.com/introducing-support-for-local-ai-models-in-the-antigravity-sdk/) Google Developers Blog Local AI Agents

### [Introducing Support for Local AI Models in the Antigravity SDK](https://developers.googleblog.com/introducing-support-for-local-ai-models-in-the-antigravity-sdk/)

Build offline, privacy-first agents with the Antigravity SDK, running Gemma 4 26B A4B locally through LiteRT-LM, with hybrid cloud-and-local orchestration.
Read blog post → [![Bring State-of-the-Art Agentic Skills to the Edge with Gemma 4](https://developers.google.com/static/edge/litert/images/landing/gemma4_banner_2.png)](https://developers.googleblog.com/bring-state-of-the-art-agentic-skills-to-the-edge-with-gemma-4/) Google Developers Blog Agentic AI

### [Bring State-of-the-Art Agentic Skills to the Edge with Gemma 4](https://developers.googleblog.com/bring-state-of-the-art-agentic-skills-to-the-edge-with-gemma-4/)

Deploy agentic and multi-step planning capabilities entirely on-device with the new Gemma 4 family and LiteRT.
Read blog post → [![High performance generative language model inference on device](https://developers.google.com/static/edge/images/edge-run-llm-on-device.jpg)](https://developers.googleblog.com/blazing-fast-on-device-genai-with-litert-lm/) Google Developers Blog Generative AI Runtime

### [Blazing Fast On-Device GenAI with LiteRT-LM](https://developers.googleblog.com/blazing-fast-on-device-genai-with-litert-lm/)

Deploy Gemma 4 using LiteRT-LM with
Multi-Token Prediction for up to a 2.2x speedup, dynamic loading, and
built-in Swift and WebGPU APIs.
Read blog post → [![Mastering Edge AI on Raspberry Pi with LiteRT and Gemma](https://developers.google.com/static/edge/litert/images/landing/rpi-gemma-banner.png)](https://developers.googleblog.com/mastering-edge-ai-on-raspberry-pi-with-litert-and-gemma/) Google Developers Blog Edge AI \& Robotics

### [Mastering Edge AI on Raspberry Pi with LiteRT and Gemma](https://developers.googleblog.com/mastering-edge-ai-on-raspberry-pi-with-litert-and-gemma/)

Deploy secure, real-time edge AI on Raspberry Pi 5 with LiteRT. Run Gemma models on CPU using XNNPACK, YOLO on VideoCore VII GPU, and orchestrate agentic workflows using the LiteRT CLI.
Read blog post → [![Google AI Edge: Small Language Models, Multimodality, and Function Calling](https://developers.google.com/static/edge/images/edge-article-4.jpg)](https://developers.googleblog.com/google-ai-edge-small-language-models-multimodality-rag-function-calling/) Google Developers Blog SLM \& Multimodal

### [Google AI Edge: Small Language Models, Multimodality, and Function Calling](https://developers.googleblog.com/google-ai-edge-small-language-models-multimodality-rag-function-calling/)

Explore the latest techniques in Retrieval-Augmented Generation (RAG), multimodal vision-language models, and function calling for edge language models.
Read blog post →

## Web \& cross-platform deployments

[![LiteRT.js web AI client for running models in the browser](https://developers.google.com/static/edge/litert/images/landing/LiteRTjs.png)](https://developers.googleblog.com/litertjs-googles-high-performance-web-ai-inference/) Google Developers Blog Web AI \& WebGPU

### [LiteRT.js: Google's High-Performance Web AI Inference](https://developers.googleblog.com/litertjs-googles-high-performance-web-ai-inference/)

Run high-performance ML and GenAI models in the browser using WebAssembly, WebGPU, and WebNN with the official LiteRT.js library.
Read blog post →

## Community \& technical guides

[![Edge AI from the Trenches: A Practical Guide to LiteRT](https://developers.google.com/static/edge/litert/images/landing/LiteRT_Blog4.jpg)](https://dev.to/jdshah/series/39748) Dev.to Series In-Depth Guide

### [Edge AI from the Trenches: A Practical Guide to LiteRT](https://dev.to/jdshah/series/39748)

A comprehensive multi-part community guide exploring the architectural differences between LiteRT and LiteRT-LM, quantization strategies, and production deployments.
Explore series →

## Stay connected

[![GitHub discussion forums and source code repository](https://developers.google.com/static/edge/litert/images/landing/github.svg)
GitHub Discussions \& code Collaborate with the community, explore source code, and report issues. google-ai-edge/LiteRT →](https://github.com/google-ai-edge/LiteRT) [![Hugging Face community model repository](https://developers.google.com/static/edge/litert/images/landing/huggingface-color.svg)
Hugging Face Hub Explore and download pre-converted LiteRT community models. LiteRT Community Models →](https://huggingface.co/litert-community) [Feedback \& requests Share feedback and feature requests to help shape the LiteRT roadmap. LiteRT Feature Intake →](https://forms.gle/xXoDqwPYN9Q9NhYBA)