Live catalog · verified specs · real open-source models

Find the right NVIDIA hardware to run open-source AI — explained for beginners.

Natalie's AI Hardware Co. matches real, currently-shipping NVIDIA GPUs and AI servers to real open-source large language models, and shows you exactly why — in plain English, with the math shown.

12
Real NVIDIA products
5
Verified open models
4
Hardware tiers
0
Logins required
How it works

From "which model?" to "what do I buy?" in four steps.

No hardware background needed — every number on this site is explained the moment you hover or tap it.

  1. Pick an open-source AI model from our catalog — each one is a real model with a verified permissive license and public parameter count.
  2. We calculate how much GPU memory it needs using a simple, transparent formula: parameters (in billions) × 1 GB, plus 20% overhead.
  3. We match that requirement to real hardware — the cheapest setup that actually fits, or a scalable production setup if you're planning for real customer traffic.
  4. You see the full picture: total cost, total power draw translated into homes and EV batteries, and — if you need multiple machines — a plain-English explanation of what a "cluster" actually is.
Explore

Three ways into the platform

Why this matters
Choosing AI hardware isn't just "buy the biggest GPU." Memory, power, and cost scale together — a model that needs 800GB of memory can't run on a card with 32GB, no matter how fast that card is. This platform exists to make that relationship visible before you spend money.
Try the Interactive Comparison Tool →