Compare
A · run_307daavsB · run_a11225code-reviewSwap A and B- Duration
- 22.9s -41%
- A: 38.7s
- Cost
- $0.021 -71%
- A: $0.073
- Tokens
- 10.1k -62%
- A: 26.5k
- Steps
- 11 +57%
- A: 7
- Model changed
- Gemini 2.5 Pro
- A: GPT-4.1
Steps, aligned
| Change | Step | A | B |
|---|---|---|---|
| Faster | Input policy check | 0.7s | 0.2s |
| Faster | Plan the task | 7.1s | 4.0s |
| Faster | search_codebase | 2.2s | 0.8s |
| Faster | Retrieve context · top 8 | 2.7s | 1.0s |
| Faster | Decide next step | 9.1s | 2.7s |
| Faster | run_tests | 2.9s | 1.1s |
| Changed | search_codebase | 11.6s | 2.1s |
| Added | Decide next step | — | 3.2s |
| Added | get_diff | — | 0.8s |
| Added | run_tests | — | 1.4s |
| Added | Write final answer | — | 4.2s |
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.