Fine-Tune Cost Lab

Method

How the prices are collected

Sources, units, checks and limits, in plain terms.

  1. Source. Each price is read from the provider's own public pricing page. We store the page address, the date and time we read it and the exact snippet of text that holds the price.
  2. Unit. Everything is USD per GPU per hour, before tax. Multi-GPU configurations are divided down to one GPU and the total is shown beside it.
  3. Checks. A price must be between $0.05 and $150 per GPU-hour, its digits must appear in the stored snippet, and the same configuration is never listed twice. A row that fails any check is dropped, not shown.
  4. What is left out. Spot, reserved and negotiated prices, storage, network egress and tax. These are list prices, not what a given customer pays.
  5. Freshness. Every row shows the date it was read. The latest read of the stored data is 12 November 2026. Runcrate's own prices come live from runcrate.ai each time a page is built.
  6. Order. Tables sort by price, lowest first. Nothing is pinned, sponsored or boosted. Runcrate is a provider like the others and is marked as ours.

Found a wrong number? Tell us.