Kept from a run that tested Amazon Nova Pro (view run). The cases stay encrypted and are deleted on the date above, when you delete the test set, or when sampling is turned off.
Replays these cases on a cheaper candidate with the same judge and quality checks. Results stay on this test set and do not change swap recommendations, guardrails, or rollout gates.
Demo workspaces cannot run re-checks.
| Candidate | Result | Started | Details |
|---|---|---|---|
| Claude Haiku 4.5 on Bedrock (regional profile) | 29 of 30 equivalent | Nov 8, 12:23 AM | View run |