Your first local AI experiment
A local model is a language model running on your computer. A useful first experiment is a short conversation in LM Studio, before you try connecting a model to an app.
1. Start with the requirements
Read LM Studio’s current system requirements before installing. Available memory, model size, and the model’s context settings affect what will fit. Don’t assume that every model with “small” in its name is a good match for your Mac.
2. Download one model
Install LM Studio from its official website. Use its model discovery interface to choose a model appropriate for your system. Read the model card and license, and keep the initial experiment small. Installation and downloading models require a network connection.
3. Ask a repeatable question
Load the model and try: “Explain a SwiftUI view to a beginner in three sentences.” Save the exact prompt and model identifier. Then ask a follow-up. Observe response quality, responsiveness, and memory use rather than treating a fast answer as a correct answer.
4. Understand the privacy boundary
A locally executed conversation can stay on your device, but that does not make every feature, integration, or model download offline. Review settings before using personal material. Start with non-sensitive sample text, and verify generated code before running it.
Next experiment
Compare two models with the same prompt and settings. Record the differences. That is the start of a useful benchmark—not a single impressive screenshot.
Sources: LM Studio getting started · Official documentation and requirements.