Windsurf Model Benchmark: Which AI Builds a Vue TODO App Fastest?

I asked Claude, GPT-4.1, Gemini, SWE-1, and more to build the same Vue + Pinia TODO app in Windsurf. The results? Most models used outdated patterns, struggled with folder creation, and bloated global CSS — despite producing working apps

Windsurf offers many models, and it’s sometimes hard to choose what is the best. Today I’ll show you results of a test, when I ask these models to create a simple Vue application with Pinia store showing TODO list, and with ability to add new items and delete them.

In the test I measured time from hitting enter until starting the server. So it includes time to accept console commands, installing dependencies, and starting the server. While these are mostly same, they didn’t change comparable results in this set. I didn’t exclude these to show realistic time spend on creating the app from user perspective.

The prompt isn’t complicated:

In folder xxx create a Vue app with
Pinia that shows a TODO list. User should
be able to add items, and delete them

Claude 3.7

The model crashed at this command: