Requesting guidance on Google Tensor SDK access for Pixel 10 NPU development

Google AI Edge
1

Hi Google Developer team,

I have already submitted an access request through the website for the Google Tensor SDK, but I’m hoping to understand whether there is any way to get a faster response or confirm the
status of my request.

I’m working on running a custom ML model locally on a Pixel 10 device, specifically targeting the on-device Google Tensor NPU/TPU path rather than CPU or GPU fallback. My goal is to
evaluate model deployment through the official Google Tensor SDK, LiteRT, or any supported Google AI Edge runtime that can target the Pixel 10 NPU.

Could you help clarify:

  1. Is there an expected review timeline for Google Tensor SDK access requests?
  2. Is there a recommended way to follow up on an existing application?
  3. Are there any partner, early-access, or developer programs for Pixel 10 NPU testing?
  4. Is LiteRT currently the recommended public path for deploying custom models to Google Tensor NPU hardware?
  5. Are there public examples or documentation for compiling and deploying a custom model to Pixel 10 NPU?
  6. Where can developers find the supported operator list, precision support, and fallback behavior documentation?

This is for custom model deployment and performance validation on Pixel 10 hardware. I’m not looking for Gemini Nano or AICore managed inference specifically; I want to understand the
supported developer path for running my own model on the NPU.

Any guidance on how to proceed or how to accelerate the access request would be greatly appreciated.

Thanks.