gradientsmith
← leaderboard

claude-opus-5

200 rollouts

pass@1 clean
94%
pass@1 adversarial
89%
robustness gap
5%
p50 latency
2400 ms
p95 latency
5200 ms
cost per attempt
$0.00950
cost per solved clean
$0.0101
cost per solved adv
$0.0107

Per-task breakdown

taskfamilycleanadversarial
code/merge-intervals/001codepasspass
code/max-subarray-sum/001codepasspass
code/roman-to-int/001codepasspass
structured/receipt-cents/001structuredpasspass
structured/apache-log-extract/001structuredpasspass