Windows Local LLM - Ollama + qwen3-coder:30b

If you got a decent Nvidia GPU for gaming, you can run a local LLM to help code instead of being limited by Cursor / Antigravity free tier limits or paying monthly.

And chances are if you got a GPU for gaming, you're running windows. So here's how I went about setting it up in Windows 11.

1. Install Ollama - There is a Download button on the home page https://ollama.com/
   Since I'm on windows it auto detected and linked to https://ollama.com/download/OllamaSetup.exe. Install Ollama from the .exe file

2. Download the model via Ollama  The specific one I used was https://ollama.com/library/qwen3-coder:30b

   

   When it's ready, you should be able to ask it questions here itself or say hi to make sure it's running.

Using the model

I tried various IDEs / VSCode extensions to get something as close to Cursor as I could. It's not quite as polished as Cursor but ultimately I settled on VSCode  + Continue.dev extension.

IDEs

All IDEs and extensions I tried I was using qwen3-coder:30b.

• PearAI was continuously getting stuck in a loop. I didn't spend much time trying to tweak configurations nor other models as I specifically wanted to use qwen3

• Void worked out of the box after selecting a model but it only shows the preview code edits in the sidebar instead of inline in the file like Cursor does. You still need to click apply or use a keyboard shortcut for it to insert the changes.

VSCode Extensions

• Roo Code was getting stuck in a loop as well using the out of the box Code role. Despite some effort to create custom modes with different Role definitions, it was still getting stuck in a loop.

• Ollama Autocoder - I could not get this working well. I had to run Ctrl + Shift + P to trigger the Autocomplete with Ollama and it honestly didn't do a very good job if it did anything at all

• Continue.dev - this is similar to Void in that the changes are shown in the sidebar and you can apply the changes to the file with a click or keyboard shortcut.

VSCode + Local LLM + Continue.dev Configuration

• In the Continue extension settings, you will need to update the local config

  and the models to use:

  Select local config and generate!

  

  It doesn't do as good of a job as Cursor but that could potentially be improved by updating the agent system rules or by maybe trying other models.