SWIFT gpi: what it fixed and what it left

SWIFT gpi is the correspondent banking system’s answer to its oldest complaint: that a cross-border payment disappears into the network and nobody can say where it is. Introduced in 2017, gpi does not change how money moves. It changes what can be seen and what member banks promise about the journey, and by SWIFT’s own figures it has made that journey much faster than its reputation suggests.

What gpi actually is

gpi is a service layer on the SWIFT messaging network, built on three commitments that member banks sign up to.

First, tracking. Every gpi payment carries a UETR, a unique end-to-end transaction reference that stays with the payment across every hop. The sending bank, and through it the paying customer, can see each leg of the journey: which bank holds the payment, when it was passed on, and when the beneficiary was credited.

Second, service levels. gpi members commit to same-day processing of gpi payments received within their business hours, and to sending a confirmation when funds reach the beneficiary’s account.

Third, transparency. Members disclose the fees they deduct and the FX rates they apply, so the sending bank can reconstruct what happened to the amount in flight rather than discovering it from an unhappy beneficiary.

What it fixed

The published numbers are genuinely strong. Nearly 60% of gpi payments are credited to the end beneficiary within 30 minutes, almost 100% within 24 hours, and member banks send over $300 billion a day through the service. The old picture of the week-long international payment is, on gpi corridors, out of date.

The tracking data also corrected a myth about the network itself. A BIS analysis of gpi flows found that payments cross fewer intermediaries than commonly assumed, and that the in-flight legs between banks are usually quick. Most of the elapsed time accrues at the receiving end, in the beneficiary bank’s opening hours, batch cycles, local practices, and checks, which is where cut-off times and time zones do their compounding work.

What it left

Read the commitments carefully and the boundary of gpi becomes clear: every one of them is about the message, the observation, and the behavior of members. None of them is about the settlement model.

A gpi payment still moves the way correspondent payments have always moved, by debits and credits across nostro and vostro accounts held between banks in the chain. That means the structural costs remain where they were. Banks still pre-fund accounts in each corridor to make payments possible, and that parked capital is a cost no tracking layer can release. Intermediaries can still deduct fees in flight; gpi makes the deduction visible rather than absent. And a payment that misses a cut-off or lands in a compliance queue is precisely tracked while it waits.

Tracking a payment, in other words, is not the same as changing what a payment is. gpi holds a mirror up to the correspondent chain and disciplines its members, which is real progress. The chain itself, with its accounts, its funding, and its stacked charges, is what the search for alternatives to SWIFT is actually about.

Where Frame fits

Frame is the settlement layer for global finance: one integration to orchestrate payments at scale across fiat rails, stablecoins, and tokenized deposits, with compliance enforced on every transaction and settlement in seconds instead of days.

gpi and a settlement layer answer different questions. gpi tells you where a payment is inside the correspondent chain; a settlement layer chooses which rail the payment should take at all, correspondent network included, and enforces the policy that governs it inside settlement itself. Frame’s Rules Engine evaluates every transaction against its governing policies before it settles, and every settled transaction produces verifiable evidence that its conditions were met. Visibility comes from the design, rather than from a tracking layer added afterwards.

See how a rail-neutral settlement layer works: the Frame Blueprint.