Compare
A · run_96a8a6vsB · run_611795code-reviewSwap A and B- Duration
- 11.3s +253%
- A: 3.2s
- Cost
- $0.056 +180%
- A: $0.020
- Tokens
- 20.5k +8%
- A: 19.0k
- Steps
- 7 -36%
- A: 11
- Model changed
- GPT-4.1
- A: Claude Haiku
Steps, aligned
| Change | Step | A | B |
|---|---|---|---|
| Slower | Input policy check | 0.0s | 0.2s |
| Slower | Plan the task | 0.3s | 2.7s |
| Slower | search_codebase | 0.3s | 1.4s |
| Slower | Retrieve context · top 8 | 0.2s | 1.0s |
| Slower | Decide next step | 0.4s | 1.8s |
| Slower | run_tests | 0.2s | 1.0s |
| Changed | Write final answer | 0.2s | 2.6s |
| Removed | Decide next step | 0.6s | — |
| Removed | get_diff | 0.2s | — |
| Removed | run_tests | 0.1s | — |
| Removed | Write final answer | 0.5s | — |
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.