---
title: "Discussion on \"A lightweight model router for Gemini agents on CPU-only Cloud Run\" | Hashnode"
description: "Discussion on \"A lightweight model router for Gemini agents on CPU-only Cloud Run\". Cut Gemini agent costs on Cloud Run using Laya, a lightweight 421M model that routes requests on CPU without GPUs or third-party APIs."
url: "https://hashnode.com/posts/a-lightweight-model-router-for-gemini-agents-on-cpu-only-cloud-run/6add87eef31315eb13cc302d"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [A lightweight model router for Gemini agents on CPU-only Cloud Run](https://derrickqin.com/a-lightweight-model-router-for-gemini-agents-on-cpu-only-cloud-run?utm_source=hashnode&utm_medium=feed)

Cut Gemini agent costs on Cloud Run using Laya, a lightweight 421M model that routes requests on CPU without GPUs or third-party APIs.

Oct 25 · 8 min read · derrickqin.com  
Read the article

-   [#cloud-run](https://hashnode.com/tag/cloud-run)
-   [#laya](https://hashnode.com/tag/laya)
-   [#jev](https://hashnode.com/tag/jev)
-   [#adk](https://hashnode.com/tag/adk)
-   [#agent](https://hashnode.com/tag/agent)

## Written by

[Derrick Qin](https://hashnode.com/@derrickqin)

@derrickqin · Melbourne, Victoria, AU

  
[View profile](https://hashnode.com/@derrickqin)

Senior Cloud Data Architect on AWS and GCP

## Responses

Markdown supported · ⌘↵ to post

No responses yet.