Transparent product evidence

QuoteMatch internal benchmark

Run on August 23, 2026 (UTC) using the production matching configuration. These results describe a bounded 20-case test set—not guaranteed accuracy for every catalog or inquiry.

20/20Expected SKU results
1.88sAverage latency
2.86sP95 latency
4Languages tested

What was tested

English, Malay, Spanish and French inquiries; aliases; spelling errors; multi-product requests; explicit SKUs; stock conflicts; and products that should remain unmatched.

Safety design

Prices and stock values never come from the model. The model may select only a SKU from the supplied catalog. Totals are calculated in deterministic code, and uncertain, unknown or out-of-stock results require human review.

Important: Catalog quality affects matching. Always review product identity, quantity, price, stock, discount and tax before sending a commercial quote.

Try the sample · Developer quick start · Return to QuoteMatch