Answers to the common questions about billing, accounts, data, and reading results.
You fund a USD wallet and that balance is your spend cap. Each job debits the actual cost of the compute it consumes. There is no subscription, no seat license, and no fixed tier. Design count is unlimited, and the whole pricing surface is one number, your balance.
Yes. The minimum top-up is $20, and that balance is your spend cap. There are no fixed packages, so you can top up any amount at or above the minimum.
No. Design count is unlimited and there is no per design charge. Your wallet balance is the only ceiling. A large request fans out into a campaign that drains the wallet as compute is delivered and pauses if the balance runs low, so everything already produced stays downloadable.
When you submit a job, a hold is placed against your wallet for the estimated cost. The job is billed for actual GPU minutes consumed, up to the estimate. If the job ends up cheaper than estimated, the surplus returns to your wallet automatically.
Cancellation stops the run but charges for the GPU minutes already consumed. There is no refund for compute we already paid for. A job cancelled before the first heartbeat reports zero GPU consumed settles at zero and the full hold returns to your wallet.
No. There are no subscriptions, seats, tiers, or credits, and no minimum monthly spend. You pay only for the compute your jobs use. Top ups are final and unused balance does not expire.
It depends on the cause. If the failure is on our side (Modal pod crash, our tool image bug, our preflight letting bad input through, no progress timeout), the full hold is refunded. If the job completed normally but produced zero passing designs, if you cancelled it mid run, or if our overrun safety system stopped a runaway job, you pay for the GPU consumed. Ambiguous cases default to refund.
The full billing model and the safety guardrails that prevent surprise bills are on the pricing page.
Sign up with your email at tools.ranomics.com. New accounts start with a small wallet balance so you can verify a tool works on your inputs before topping up.
Your uploads and results belong to your account and are tied to the email you signed in with. A job link opened under a different account will not show another user's results. See the privacy policy for the full detail.
Your inputs are validated before any compute runs, so a bad chain ID or a malformed PDB is caught up front rather than after you have paid. Your $5 signup balance also covers a small first run to see real output on your target.
It depends on the tool and the scale you pick. A single ProteinMPNN inference finishes in about a minute. Generative tools take longer as you request more designs, often a minute or two per design, so a small starter run lands well under an hour while a large campaign fans out across GPUs to finish sooner. Every tool form shows a typical runtime before you submit.
A design is one candidate the tool generates, a backbone with a sequence, scored and ranked against the others in the run. A pilot of about 8 designs gives a real first signal because roughly 1 in 5 passes the in-silico filter on a tractable target.
ipTM is interface confidence, higher is better, aim above roughly 0.7 on a tractable target. pLDDT is per-residue fold confidence, higher is better. i_pAE and pAE are predicted alignment error, lower is better.
From a finished job you can open any candidate in the 3D viewer and export the score table as CSV or the sequences as FASTA. Star the best candidates to build a shortlist for the wet lab.
Each per-tool guide explains what the tool is built for and how to read its output. The catalog compares them side by side.
If the docs do not answer it, email the team and we will help.
