Models & hardware
Find a model that fits.
A conservative starting point for Ollama on your Mac or Windows PC. No hardware scan runs from this page.
1. Check your hardware
Open Terminal on Mac or PowerShell on Windows. Run the matching read-only command, then paste its single compact line below. These commands report CPU, memory and GPU details only; they do not install or change anything.
Mac: show command
Reads the chip or CPU name and installed memory; Apple silicon uses unified memory.
Windows: show command
NVIDIA memory is read with nvidia-smi when available. AMD, Intel, missing drivers or unknown VRAM use the CPU fallback. Windows AdapterRAM is not used because it can misreport larger cards.
Your pasted output stays in this page's memory. It is not sent to a server, saved to browser storage, added to a URL, logged or passed to analytics. Refreshing clears it.
Or choose manually
For Intel Macs choose CPU only. For multiple NVIDIA cards choose the memory of one card; capacities are not added together. The 48 GB choice is a conservative floor for larger cards.
How this estimate works
Apple silicon uses 70% of unified memory. NVIDIA uses VRAM minus at least 2 GB or 20%, whichever is larger. CPU fallback reserves at least 4 GB or half of system RAM. These are Alluqi planning assumptions, not official model requirements. More installed memory does not guarantee a better experience.
Model sizes checked 2026-10-27. Exact tags and sizes can change; review the official library before downloading. Downloading models uses network data and disk space.