Compare
A · run_611795vsB · run_96a8a6code-reviewSwap A and B- Duration
- 3.2s -72%
- A: 11.3s
- Cost
- $0.020 -64%
- A: $0.056
- Tokens
- 19.0k -7%
- A: 20.5k
- Steps
- 11 +57%
- A: 7
- Model changed
- Claude Haiku
- A: GPT-4.1
Steps, aligned
| Change | Step | A | B |
|---|---|---|---|
| Faster | Input policy check | 0.2s | 0.0s |
| Faster | Plan the task | 2.7s | 0.3s |
| Faster | search_codebase | 1.4s | 0.3s |
| Faster | Retrieve context · top 8 | 1.0s | 0.2s |
| Faster | Decide next step | 1.8s | 0.4s |
| Faster | run_tests | 1.0s | 0.2s |
| Changed | search_codebase | 2.6s | 0.2s |
| Added | Decide next step | — | 0.6s |
| Added | get_diff | — | 0.2s |
| Added | run_tests | — | 0.1s |
| Added | 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.