Benchmark

Claude Opus 5.5 in Antigravity: TwentyDollarAI Standard Benchmarking

Interrupted by the five-hour limit. Usage notes, plus the output.

Results

Time is my estimate. Quota from my Google Pro usage meter.

Measured results: metric and value
Time (total runtime)About 24m
Tokens (total)Not measured
Cost (API-equivalent)Not measured
Five-hour limitHit mid-test
Weekly usageShared (see notes)

Models tested

  • Claude Opus 5.5Plan: Google Pro plan

The prompt

Standard TwentyDollarAI Benchmarking

Evidence

Watch the toy: a glowing yellow blob, not my dog's chicken. The tooltip and cursor in the first second are from my screen recording, not the animation.

My take

  • Attractive scene: recognizable Maltese, rich meadow.
  • Weaker fidelity: toy and dog less true to the reference.
  • Limits matter: testing paused. Scores and notes