$ replaybook

Model results

Claude Sonnet 5.5

anthropic/claude-sonnet-5.5

Each row is an independent scenario cohort. Releases, providers and reasoning settings remain separate; these results are not pooled into a single model score.

Repairs, timing and cost

ScenarioReleaseProvider / reasoningRepairsMedian timeKnown costEvidence
Reservation write contention v1frontier 20261029.0.0OpenRouter / high 3/30 unavailable 44s $0.23703/3 trials report cost Inspect attempts →
Notification outbox rollout v1frontier 20261029.0.0OpenRouter / high 3/30 unavailable 34s $0.21263/3 trials report cost Inspect attempts →
Checkout dependency rollout v1frontier 20261029.0.0OpenRouter / high 3/30 unavailable 33s $0.21843/3 trials report cost Inspect attempts →

Compare benchmark configurations → · All models