Ant International has introduced Falcon Time-Series Transformer (TST) AI Model 2.0, its most advanced TST model so far designed to deliver more accurate forecasting in real-world FX risk management of cross-border payments, with more industry applications to come, such as demand forecasting for supply chain management for e-commerce platforms, and predictive operations management for aviation industry.

FalconTST 2.0 demonstrates State-of-the-Art (SOTA) performance on the Mean Absolute Scaled Error (MASE) metric on a top global public evaluation benchmark for time-series foundational models. MASE is among the most critical metrics used to evaluate time-series models. FalconTST 2.0 achieved a MASE score of 0.666 and places it at the top of the leaderboard, surpassing other TST foundational models from leading global tech companies.

FalconTST model is AI built for finance

While large language models excel at learning relationships in text, TST models are especially critical in finance and payments, where liquidity needs, foreign-exchange movements, and transaction flows can shift rapidly. The financial information consists of continuously changing numerical data — transaction amounts, account balances, settlement flows, and currency positions. For a global payment institution, these forecasts directly impact capital efficiency. The value of AI prediction lies not just in ‘calculating more accurately’ but in helping businesses know precisely when they need funds, how much they need, and in which currencies.

This forecasting capability is equally critical for foreign exchange management. An airline may collect ticket revenues in multiple currencies while needing to pay for aircraft leases, airport fees, and operating costs in different currencies. Companies typically use foreign exchange hedging to reduce currency fluctuation risk, but that requires them to determine how much of each currency they will receive and need. If forecasts are too high, they may over-hedge; if too low, they leave themselves exposed to foreign exchange risk.

Traditional forecasting systems typically build separate models for different tasks — a retail company trains a sales forecast model, an airline a demand forecast model, a financial institution a liquidity model. TST foundational models take a different approach: FalconTST learns common patterns — cycles, trends, seasonality, and sudden shifts—from data across finance, retail, energy, travel, and economics. Though these industries differ, the underlying temporal structures often share commonalities.

World’s leading banks using FalconTST to epitomise liquidity and FX management

The FalconTST is initially deployed internally at Ant International to manage cashflow and FX exposure on an hourly, daily and weekly basis, before having been integrated by leading global banks to their own FX hedging models, including Barclays, Citi, Deutsche Bank and Standard Chartered, to improve the cashflow forecasting and FX liquidity management capabilities for Ant International and its clients.

Barclays integrates the FalconTST Model into its FX hedging platform, BARX NetFX, while Citi combines FalconTST with their own Fixed FX Rates solution. They are mainly used for FX risk management on e-commerce platforms or airlines. And Standard Chartered uses the model alongside its SCALE FX system as part of both sides' participation in the PathFin.ai programme of the Monetary Authority of Singapore.

Currently, they all have adopted the 2.0 version of the FalconTST Model, leading to an improved forecasting accuracy of more than 93% consistently. This level of precision is critical for financial institutions managing vast volumes of cross-border payments and needing to mitigate currency fluctuation risks effectively.