Build my AI machine
Start with the work, not a product name. Results use verified specification-sheet fields only. They do not use live prices, listings, accounts or stored visitor data.
Your machine shortlist
Choose a workload to begin.
Stack for this build
Software stacks are coming to this page.
How we calculate
Memory fit: model parameters in billions × bytes per parameter (FP16 2.0, Q8 1.0, Q4 0.55 including weight overhead), plus the selected context reserve, then 15% headroom. A candidate is Tight when verified memory clears the requirement by less than 20%; below the requirement is Won't fit.
Scoring: memory fit leads. Fine-tuning adds weight for verified memory and bandwidth class; image generation and video add weight for verified compute tier; agent and RAG choices add memory/context weight. Results are sorted, deduplicated by machine class and limited to three.
Budget: filters catalog classes (entry, mid, upper, flagship); it never reads or claims a live price.
Models missing a required field are excluded and never estimated.
Guidance from our verified spec sheets — verify fit for your exact model and software before purchase.