Compare
A · run_4f67e0vsB · run_85d1cacode-reviewSwap A and B- Duration
- 21.9s -44%
- A: 38.8s
- Cost
- $0.015 -70%
- A: $0.049
- Tokens
- 12.0k -34%
- A: 18.2k
- Steps
- 8 +14%
- A: 7
- Model changed
- Claude Haiku
- A: GPT-4.1
Steps, aligned
| Change | Step | A | B |
|---|---|---|---|
| Faster | Input policy check | 0.7s | 0.3s |
| Faster | Plan the task | 10.8s | 3.8s |
| Faster | search_codebase | 3.9s | 1.5s |
| Faster | Retrieve context · top 8 | 3.8s | 1.7s |
| Same | Decide next step | 7.4s | 5.1s |
| Faster | run_tests | 3.1s | 1.7s |
| Changed | search_codebase | 6.8s | 1.9s |
| Added | Write final answer | — | 4.5s |
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.