Hardware cheat sheet: what you can realistically run at each RAM tier

Here's a practical guide to what local AI hardware can do. With 8 GB of RAM, start small and keep your first test simple. With 16 GB, you can run useful experiments with models like E4B and smaller quantized models. With 32 GB, you have room for larger local workflows. With a strong GPU or a workstation like a DGX Spark, the bigger models become much more realistic. For phones, think less about model size and more about the job: Can the model understand a photo? Can it summarize audio? Can it classify something quickly? Can it help a field worker without a strong connection? Can it do something useful inside the app before the user even thinks to ask?