§ 001 / GPU BUY CHECK

Before you buy a GPU for local models, have a person check the plan.

You tell me your current card, your budget, the GPUs you are weighing and the models you want to run. I send back one written answer within 5 days:

• Buy, do not buy, or buy this instead.
• The tokens per second and VRAM headroom to expect for your models.
• The llama.cpp or Ollama settings to start with.
• Which numbers come from measured runs on our own rigs, and which are estimates.

$49, paid once. If the answer does not help, you get a full refund.

No call and no ongoing help. You fill one form, and I write back once. I take 3 orders a week, so the page says when the week is full.

Want the free answer first? The sizing desk shows what fits your card today.