Loading page…
The model targets a specific treasury problem in cross-border payments: predicting the timing, size and currency of cash needs. Its reported bank integrations are meaningful, but their scale—and one headline accuracy measure—remain unclear.
Listen to this story
FalconTST 2.0 is moving from demonstration to targeted financial workflows: Ant International says four named banks are using it for treasury forecasting, FX liquidity, or hedging, while six banks have signed partnerships. Ant reports a 0.666 Mean Absolute Scaled Error on a public benchmark and accuracy above 93%, but does not identify the latter benchmark.
Ant announced FalconTST 2.0 on August 19 for sequence data such as transactions, balances, settlements and currency positions.
Named deployments are at Barclays, Citi, Deutsche Bank and Standard Chartered; HSBC is among the six partnership signatories.
Ant’s 0.666 benchmark score is more comparable than its separate “above 93%” accuracy claim, whose benchmark is undisclosed.
Ant International says its FalconTST 2.0 forecasting model is now being used in tools at Barclays, Citi, Deutsche Bank and Standard Chartered for cash-flow forecasting and foreign-exchange liquidity or hedging management. The deployment puts AI into decisions over how much cash to prepare for cross-border payments and which currencies to hold.
The company announced Falcon Time-Series Transformer 2.0 on August 19. It is built to forecast foreign-exchange risk and cash-flow needs for financial institutions, rather than to handle the broad text tasks associated with general-purpose chatbots.
Ant describes the relevant information as numbers that keep changing: transaction amounts, account balances, settlement flows and currency positions. A time-series model is designed for that kind of sequence data, where the recent path of values is part of the forecast.
Ant says the useful output is not simply a better score. The forecast is meant to help a business determine when it needs funds, how much it needs and in which currencies—inputs to preparing liquidity and managing FX exposure.
Ant International said FalconTST 2.0 achieved a 0.666 Mean Absolute Scaled Error score on a public evaluation benchmark.
Ant says six major banks have signed partnerships involving FalconTST 2.0. It named Citi, HSBC, Deutsche Bank, Standard Chartered and Barclays, leaving one partner unnamed. The company separately says Barclays, Citi, Deutsche Bank and Standard Chartered are using the model.
The public description identifies the functions receiving the model, not the breadth of each bank’s use. Ant also reported forecast accuracy above 93%, but did not identify the benchmark behind that figure. Those two limits make the announcement evidence of specific integrations, not a measure of enterprise-wide adoption or a fully comparable performance claim.
Ant says FalconTST 2.0 was already being used for foreign-exchange and liquidity management in aviation. It plans to extend the model to e-commerce supply-chain demand forecasting and predictive operations management for aviation, applying the same kind of forecasting to different streams of business data.
The company has also invited developers to an API trial through GitHub. That opens a route for outside developers to test applications beyond Ant’s initial partners, while the bank announcement remains centered on treasury-related tools.
The next concrete test is whether named banks expand FalconTST 2.0 beyond the cash-flow, liquidity and hedging tools Ant has described.
Loading discussion...
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.