Compare
A · run_fe9d90vsB · run_023c10code-reviewSwap A and B- Duration
- 18.3s -44%
- A: 32.9s
- Cost
- $0.103 +69%
- A: $0.061
- Tokens
- 28.9k +12%
- A: 25.7k
- Steps
- 5 -50%
- A: 10
- Model changed
- Claude Sonnet
- A: GPT-4.1
Steps, aligned
| Change | Step | A | B |
|---|---|---|---|
| Same | Input policy check | 0.4s | 0.4s |
| Same | Plan the task | 5.5s | 4.8s |
| Same | search_codebase | 1.9s | 2.9s |
| Same | Retrieve context · top 8 | 2.0s | 2.5s |
| Changed | Write final answer | 4.7s | 6.6s |
| Removed | run_tests | 1.4s | — |
| Removed | search_codebase | 1.6s | — |
| Removed | Decide next step | 5.4s | — |
| Removed | get_diff | 1.9s | — |
| Removed | Write final answer | 6.1s | — |
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.