Run the pilot

1. Load the target and define the hotspot sets you want to test.
2. In Plan, choose Pilot and select the generators to compare.
3. Use the same design count and campaign thresholds for every generator and hotspot set.
4. Review the cost estimate, then submit the pilot.
5. Compare the pass rate and passing count for each combination. Send the nominated combination to the batch.

The public benchmark below used fifty designs per combination. It contains 4,813 miniprotein designs on five targets from three generators, scored with AlphaFold2 at Dyno's current recommended thresholds.

Compare generators on the same target

Pick a target below. The figure shows the difference between its best and worst generator at the same thresholds.

Check the cost

Generators charge differently. A batch model produces the number you ask for; a hunt model holds a GPU until enough designs pass, so the number you ask for is a budget and a hard target can consume it. Every submit shows a cost estimate first and a dry run is free, so the pilot itself is priced before it runs.

Reuse an earlier pilot

Reuse the previous results when your organization has already piloted the same target and site. Phi keeps pilot results with the campaign so another campaign on that target can start from them.

Checked 2026-10-22