Research & models Nov 15, 2026
A JPMorgan Chase preprint reports that the Large Payments Model, a foundation model trained from scratch on payment transaction data, is serving in production at under US$0.0001 per inference.

JPMorgan reports large fraud-detection gains over existing models, but the paper discloses neither the comparison models nor absolute values, so outsiders can't check the baseline. According to the paper, average precision in merchant-acquiring fraud detection rose by between 62.7% and 85.5% relative to existing models, a relative gain rather than percentage points; 99% of serving requests return in under 60 milliseconds; and wire-fraud detection measured by amount showed no improvement. All the authors work at JPMorgan Chase, and the results come only from JPMorgan's own tests (arXiv 2611.15814, 2026-11-11). How we checked →
Source and checks
This item ran in the brief of Nov 15, 2026, alongside 9 other items. How we checked it, the quote checks and what was still missing that day are on that issue's How we checked page.
Cite this item
SecondSource (2026). "A JPMorgan Chase preprint reports that the Large Payments Model, a foundation model trained from scratch on payment transaction data, is serving in production at under US$0.0001 per inference.." https://secondsource.io/en/research/2026-11-15-2/. Accessed access date.