Yesterday, OpenAI released two new models, GPT-OSS-120B and GPT-OSS-20B, optimized for reasoning. The 20B model is particularly interesting because it’s designed to run locally, making it easier to experiment with while still accessing it programmatically via an API.
To get started, you can pull and run GPT-OSS-20B with Ollama, and then expose it as an OpenAI-compatible API using Local Runners. This setup makes it simple to test the model locally while still having the flexibility of API access.
If you’re interested in giving it a try, I’ve included a guide link in the comments below. It walks you through the process step-by-step, so you can get up and running quickly.
What are your thoughts on running AI models locally? Is this something you’d be interested in trying out?