Can I run this model on my computer?

Start with the weight-file screen

These are decimal GB limits from Finder's current rule, not measured runtime requirements. Add all the required weight shards for the selected variant before comparing. Do not compare the size of an entire repository containing several alternative quantizations with one model's weight size.

Check four things before a large download

1. Identify your system RAM and, on a discrete GPU, VRAM. Apple silicon uses shared memory.
2. Check the exact architecture, file format and companion files your runner requires.
3. Leave room for the operating system, runner and context cache. More context increases memory use.
4. Load a small test and record the exact file, runner version, context and result. A fit estimate is not execution evidence.

Use hardware pages to inspect device profiles, Finder for current files, or the 16 GB Mac comparison for a concrete starting point.

When a file does not fit

Inspect a smaller model or a smaller supported quantization, reduce context, and close other memory-heavy apps. Check common local AI problems if loading fails. Avoid treating a smaller download as proof of better quality or adequate performance for your task.