Rent your first machine

Use this route when you want to work from the website rather than write an integration.

1. Choose the work

Open Rent. Start with the model or application you need, then compare compatible offers. Check GPU count, memory, runtime compatibility, price for the whole machine, and whether the offer is Interruptible. A published template is not the same thing as a hardware-verified template.

For inference, pick the model and serving runtime you intend to call. For container work, check the template’s access type and whether you need SSH. See choosing a template and quote.

2. Sign in and add funds

Sign in at AmpleRun. Use the sign-in methods currently offered by the site. Open your account’s Add funds flow, read its supported network, asset, minimum, and address, then confirm the transfer in your own wallet or exchange.

Wait until the displayed spendable balance covers the budget. A pending transfer is not spendable money. Deposit states explain the difference.

3. Review the quote and deploy

Set a duration and a budget you are comfortable spending. Read the quote’s total, rate, included resources, fee, and interruption terms. The quote expires; refresh it if necessary. Only proceed when the workload and spending limit are correct.

A reservation progresses through RESERVED and STARTING. Access opens when the rental reaches RUNNING, after verified readiness. A start that never reaches readiness is not charged runtime.

4. Use, save, and stop

Open the access details in the rental. Use the returned connection information, not a hostname copied from another rental. Save results outside the rented machine before stopping. Runtime state and local files are not a durable backup.

Stop the rental when your work is done, then check the final state and receipt. After a completed stop, any read-only retrieval access is time-limited; inspect its expires_at rather than relying on it as storage.

Done when: the workload is stopped, your output is stored elsewhere, and the receipt matches the rental you intended to run. If anything differs, open a support case with the rental ID.