AI TOOL COMPARISON REVIEW - REVIEW KEY - VERSION 1
Keep this file out of the test conversation.

CMP-01
Preserve the two observed times and input lengths. The runs do not
establish relative speed. Use the same permitted input and task, and
keep other test conditions as comparable as possible.

CMP-02
Record could not test: the sample could not be submitted, so there
is no output-quality result. Verify the approved input path before
retrying. Do not infer a plan restriction or a content failure.

CMP-03
Keep the earlier pass as historical evidence tied to that preset.
The current preset is untested. Run the same date check with the
current preset and keep the old and new outputs separate.

Equivalent wording is acceptable if it keeps the observation separate
from the unsupported claim and names a check that could resolve the gap.