Task 1 of 88
Straighten and crop a seaside banner
Aiming
Fail
Instructions not followed. Looped without progress
- Score
- 0.40
- Instructions
- Not followed
- Time
- 40m 5s
- Model cost
- $11.49
- Tokens
- 49,943,752
- Task id
- aiming/banner_crop
Octavus Benchmarks
Overall
77
#4 of 7 overall
Per-task results
88 tasks across 8 task types, 88 with a public recording. Pick a task to see its verdict, time, and cost - and watch exactly what the model did.
Task 1 of 88
Instructions not followed. Looped without progress
Speed and cost are per task, relative to the best of the 7 models on this board.
Cost is the model / provider cost. The score is the pass rate over every task that ran; a task whose gold answer is not yet verified is still counted, and flagged "Gold unverified". Every task's trace was reviewed against the rules its instruction states: a pass that broke one is disqualified and counts as a fail, and a task that broke a rule or malfunctioned did not follow its instructions.