Video tutorial · The Metaverse Guy
Pony Alpha (GLM-5): Free AI Model for UI Coding and Auth Integration
I tested the free Pony Alpha model on OpenRouter inside Kilo Code, building a dating app UI, landing pages, and integrating Clerk auth into Next.js – all from simple prompts.
This video was made in partnership with TestSprite. It carries YouTube's paid-promotion label.
- Published
- 11 February 2026
- Length
- 11 min
- Views
- 5,502 views
Pony Alpha, also known as GLM-5, is a 100% free AI model on OpenRouter that surprised me with its coding and UI generation capabilities. In this video, I put it through its paces inside Kilo Code, a free open-source AI code editor, building real projects like a mobile dating app UI, landing pages, and 3D demos. I also ran a timed test to integrate Clerk authentication into a Next.js app, and I’ll show you how to use TestSprite for automated testing before you ship.
What is Pony Alpha?
Pony Alpha, also referred to as GLM-5, is a completely free AI model available on OpenRouter. It’s designed for coding tasks and UI generation, and in my tests it performed surprisingly well. I wanted to see if a free model could handle real-world development workflows, so I used it inside Kilo Code, a free and open-source AI code editor, to build several projects from scratch.
Setting Up Pony Alpha in Kilo Code
To get started, you need an OpenRouter account and an API key. Then, inside Kilo Code, you configure the provider to use OpenRouter and select Pony Alpha as the model. The setup is straightforward and doesn’t cost anything.
- Create an account on OpenRouter and generate an API key.
- Install Kilo Code (free, open-source AI code editor).
- In Kilo Code, add OpenRouter as the AI provider and paste your API key.
- Select the Pony Alpha model from the model list.
Demo Projects: Dating App UI and More
Once everything was set up, I tested Pony Alpha with a series of prompts. I asked it to build a mobile dating app UI, and it generated a full Expo (React Native) project from a single prompt. The output included multiple screens and components, and the code was well-structured.
I also tried landing pages and 3D demos. Each time, Pony Alpha produced working interfaces without me having to tweak the prompts heavily. It’s not just about generating code snippets – it can scaffold entire projects that you can run locally.
Real-World Test: Integrating Clerk Auth in Next.js
To push it further, I set up a fresh Next.js project and tasked Pony Alpha with integrating Clerk authentication. I used Clerk’s own generated prompt to guide the model, then started a stopwatch to time the integration.
Pony Alpha handled the task step by step: installing dependencies, setting up environment variables, and wrapping the app with Clerk’s provider. There were provider rate limits along the way, but the model managed them without breaking the workflow. After setting up the .env file and running the project, I verified that Google login worked correctly – the auth flow was fully functional.
- Create a new Next.js project.
- Use Clerk’s generated prompt to instruct Pony Alpha to add authentication.
- Start a stopwatch and let the model work through the integration.
- Handle any provider rate limits by waiting and retrying.
- Configure the required environment variables.
- Run the project locally and test the login flow.
Why Testing Matters (and How TestSprite Helps)
Building with AI is fast, but shipping without proper testing is risky. That’s where TestSprite, the sponsor of this video, comes in. It connects to your AI code editor via MCP and automatically runs frontend and backend tests. It records the sessions, reports failures, and helps you fix issues quickly.
To use it, you install the TestSprite MCP integration, generate an API key, and then trigger automated test runs from your editor. The dashboard shows you exactly what passed and what failed, so you can address problems before they reach production.
- Install the TestSprite MCP plugin in your AI code editor.
- Generate an API key from the TestSprite dashboard.
- Run automated tests on your project.
- Review the test results and fix any reported issues.
Final Thoughts
Pony Alpha is a solid free option if you want to build interfaces and handle real coding tasks without paying for a model. It’s not perfect – you’ll encounter rate limits – but for prototyping and even production-ready features like auth, it gets the job done. Just remember to test thoroughly before you ship, and tools like TestSprite can make that process much smoother.
Tools used in this video
- Pony Alpha
- GLM-5
- OpenRouter
- Kilo Code
- Next.js
- Clerk
- Expo
- React Native
- TestSprite
- MCP
Questions
- Is Pony Alpha really free?
- Yes, Pony Alpha (GLM-5) is a 100% free model on OpenRouter. You can use it without any cost, though you may encounter provider rate limits.
- How do I use Pony Alpha in Kilo Code?
- You need an OpenRouter API key. In Kilo Code, add OpenRouter as the provider, paste your key, and select Pony Alpha from the model list. Both tools are free.
- Can Pony Alpha handle real-world tasks like authentication?
- Yes, in my test it successfully integrated Clerk authentication into a Next.js app. It handled dependency installation, environment setup, and provider wrapping, and the login flow worked correctly.
- What is TestSprite and how does it help with testing?
- TestSprite is an automated testing tool that connects to your AI code editor via MCP. It runs frontend and backend tests, records sessions, reports failures, and helps you fix issues before shipping.
Topics
Chapters
- 00:00 Pony Alpha: the 100% free AI model
- 00:14 Pony Alpha on OpenRouter (context + specs)
- 00:36 Testing Pony Alpha inside Kilo Code (free code editor)
- 00:45 Projects demo: landing pages + 3D tests
- 01:28 Demo: dating app UI built from one prompt
- 02:26 How much code it generated (Expo app breakdown)
- 03:02 How to use Pony Alpha in Kilo Code / OpenRouter
- 03:30 New test: Next.js project setup
- 03:56 Task: add Clerk authentication (real-world workflow)
- 04:25 Using Clerk’s generated prompt inside Pony Alpha
- 05:02 Stopwatch test begins (timing the integration)
- 06:24 Provider rate limits (what to expect)
- 07:06 env setup + running the project
- 07:42 Clerk auth working (Google login verified)
- 08:04 Why testing matters (don’t ship blind)
- 08:24 Sponsor: TestSprite overview
- 08:42 Install TestSprite MCP + generate API key
- 09:35 Automated test runs + dashboard results
- 10:13 Fixing issues using TestSprite + your AI editor
- 10:39 Final advice + next steps
Video description and links
This article was written from the video's description with the help of AI. Watch the video for the full walkthrough.



