Credits pay for hosted computation. They are separate from the scientific scores a tool returns, and from the local computations you run with the Python SDK on your own hardware.
Review estimates before submission
Service forms expose an estimated credit cost. Runtime and credit usage can vary with the tool, system size, sampling, and settings. Treat an estimate as planning information rather than a guaranteed final charge.
Check the current form and account balance before committing to a run. Some services have additional submission requirements. For example, training can require a minimum available balance even when the exact final runtime is unknown.
Understand what changes compute use
- Larger systems and more molecular entities can require more resources.
- More generated candidates, samples, or simulation time usually increase work.
- A pipeline or bulk workflow can create many jobs; review the total expansion before submission.
- A failed or cancelled run should not be assumed to have consumed no computation. Inspect its recorded usage and the applicable billing behavior.
Start with a small, representative calculation when setting up a new workflow. Validate the input/output handoff before scaling to a library.
Check balances and recorded usage
The workspace shows account information and, when applicable, team information. Team balances are distinct from personal balances. Review which account or team the submission uses.
The jobs table exposes Credits Used for runs. Record actual usage from completed pilot runs when planning similar calculations, while allowing for differences in size or settings.
Continue when credits are low
Tool discovery and documentation remain useful even when you cannot submit a new run. Prepare inputs, compare methods, inspect previous results, and review the estimate before purchasing credits or changing your plan.
Use the workspace’s credit purchase and subscription controls for current account options. Consult current pricing rather than treating a price quoted in an old tutorial as authoritative.
Plan larger workflows
Define the number of inputs, candidates per input, stages, and filtering decisions before execution. A design stage that generates many structures can multiply the size of every downstream folding or scoring step.
For simulation planning, see Estimating molecular dynamics runtime and credits. For questions about an unexpected charge, provide the relevant job IDs when contacting support.