Compare
A · run_eea665vsB · run_ab7767invoice-extractorSwap A and B- Duration
- 39.0s +359%
- A: 8.5s
- Cost
- $0.073 +40%
- A: $0.052
- Tokens
- 10.6k -54%
- A: 23.0k
- Steps
- 10 +43%
- A: 7
- Model changed
- Claude Sonnet
- A: GPT-4.1
Steps, aligned
| Change | Step | A | B |
|---|---|---|---|
| Slower | Input policy check | 0.1s | 0.4s |
| Slower | Plan the task | 1.9s | 6.1s |
| Slower | validate_totals | 0.6s | 2.1s |
| Slower | Retrieve context · top 8 | 0.8s | 2.6s |
| Slower | Decide next step | 2.0s | 5.9s |
| Slower | ocr_pages | 0.7s | 2.4s |
| Changed | validate_totals | 1.9s | 2.0s |
| Added | Decide next step | — | 5.1s |
| Added | fetch_pdf | — | 2.7s |
| Added | Write final answer | — | 7.2s |
Output diff
− 0 words + 0 words 100% unchangedINV-2026-0412 from Maderera Sur. Total $184,320.00 with 21% VAT, due Oct 31, 2026. Line 7 did not match the PO and was sent for manual review. Posted to the ERP as a draft, held until review.