Compare
A · run_835322vsB · run_5d5c12code-reviewSwap A and B- Duration
- 15.3s -60%
- A: 37.8s
- Cost
- $0.045 +77%
- A: $0.026
- Tokens
- 18.4k -25%
- A: 24.5k
- Steps
- 6 -45%
- A: 11
- Model changed
- GPT-4.1
- A: Claude Haiku
Steps, aligned
| Change | Step | A | B |
|---|---|---|---|
| Faster | Input policy check | 0.4s | 0.2s |
| Faster | Plan the task | 4.2s | 1.9s |
| Faster | search_codebase | 2.5s | 1.4s |
| Faster | Retrieve context · top 8 | 2.5s | 1.3s |
| Faster | Decide next step | 5.7s | 2.7s |
| Changed | run_tests | 2.0s | 1.1s |
| Removed | search_codebase | 2.3s | — |
| Removed | Decide next step | 6.6s | — |
| Removed | get_diff | 3.0s | — |
| Removed | run_tests | 0.9s | — |
| Removed | Write final answer | 5.4s | — |
Output diff
− 36 words + 2 words 0% 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.(no output)