Compare
A · run_51736avsB · run_95ab4acode-reviewSwap A and B- Duration
- 12.0s -72%
- A: 43.0s
- Cost
- $0.015 -72%
- A: $0.055
- Tokens
- 15.8k -15%
- A: 18.6k
- Steps
- 6 +20%
- A: 5
- Model changed
- Claude Haiku
- A: GPT-4.1
Steps, aligned
| Change | Step | A | B |
|---|---|---|---|
| Faster | Input policy check | 1.1s | 0.2s |
| Faster | Plan the task | 13.5s | 3.0s |
| Faster | search_codebase | 9.5s | 1.5s |
| Faster | Retrieve context · top 8 | 3.6s | 1.2s |
| Changed | Decide next step | 12.7s | 3.3s |
| Added | Write final answer | — | 1.9s |
Output diff
− 0 words + 0 words 100% unchangedReviewed 14 files. Three issues: the cache key ignores the locale, a test is skipped without a reason, and a server action is missing input validation. Tests pass. Suggested changes posted as comments, one marked blocking.