Evals
Every model + prompt combination is run against a golden dataset of 30 synthetic invoices with exactly known ground truth. Scores are field-level exact-match accuracy after normalization (dates → ISO, amounts → numbers, currency → ISO 4217).
Accuracy by field
| Field | haiku-4-5 prompt v2 · 30 docs |
|---|---|
| vendor_name | 100% |
| invoice_number | 100% |
| invoice_date | 100% |
| due_date | 100% |
| currency | 100% |
| subtotal | 97% |
| tax_amount | 100% |
| total | 100% |
| line_items | 97% |
| docs 100% correct | 97% |
| avg cost / doc | $0.0047 |
| avg latency | 3192 ms |
All runs
| When | Model | Prompt | Runtime | Docs | Docs 100% correct | Avg cost/doc | Avg latency |
|---|---|---|---|---|---|---|---|
| 2026-07-10 | claude-haiku-4-5 | v2 | api | 30 | 97% | $0.0047 | 3192 ms |