# Model comparison worksheet — blank, not benchmark results

Task / audience:
Reference filename and hash:
Fixed prompt:
Acceptance criteria:
Hard rejection criteria:
Budget ceiling:

## Candidate A
Model/version:
Input mode / duration / resolution / ratio / audio:
Supported configuration verified:
Actual spend / currency:
Completed clips:
Technical failures:
Accepted clips:
Failure description and timestamp:
Editing required:

## Candidate B
Model/version:
Input mode / duration / resolution / ratio / audio:
Supported configuration verified:
Actual spend / currency:
Completed clips:
Technical failures:
Accepted clips:
Failure description and timestamp:
Editing required:

## Conclusion
Configuration differences:
Usable rate = accepted / completed (undefined if completed=0).
Cost per accepted clip = total actual trial spend / accepted (undefined if accepted=0).
Decision for this task only:
Remaining uncertainty: