Benchmark TinyStories with Google DDP and an AI agent
Who is this for?
Developers who want to use a coding agent to access remote Android hardware and compare machine learning performance on an Arm-based phone.
What will you learn?
After completing this Learning Path, you'll be able to perform the following tasks:
- Configure a Google Cloud project and local tools for Developer Device Platform (DDP).
- Verify CLI authentication, enabled DDP services, and device catalog access.
- Use the Google DDP agent skill with Antigravity to discover, reserve, inspect, and release a remote phone.
- Use Antigravity to compare TinyStories CPU inference with and without KleidiAI optimizations.
Prerequisites
Before starting, you'll need the following prerequisites:
- Access to a Google Cloud project with billing enabled and permissions to enable APIs and reserve DDP devices
- A local macOS or Ubuntu 26.04 computer with a desktop session and permission to install software
- The Google Cloud CLI and permission to install beta commands
- The Antigravity CLI coding agent
Summary
This summary was drafted with an approved AI-assisted workflow and reviewed by Arm contributors before publication. Human technical review remains part of the process so the final page reflects engineering rigor, accuracy, and Arm editorial standards.
You’ll use Google DDP and Antigravity to compare TinyStories CPU inference on a remote Arm-based phone. First, you’ll set up a Google Cloud project and local tools, then verify credentials, services, and device catalog access. After connecting to a phone, you’ll ask Antigravity to benchmark llama.cpp with and without KleidiAI. You’ll review throughput and kernel evidence, check the saved report, and confirm that the billable session has ended.