Ant International CEO Peng Yang outlined the firm’s next-stage AI priorities at the Singapore Fintech Festival, naming the Falcon TST and SHIELD models as its focus for the next 18 months.

The company has now open-sourced its Falcon TST time-series transformer model to broaden access to advanced forecasting tools.

Falcon TST is a Mixture of Experts model with multiple patch tokenisers and up to 2.5 billion parameters.

It has delivered state-of-the-art zero-shot performance on recognised long-term forecasting benchmarks.

Ant International uses the model to forecast cashflow and foreign exchange exposure on an hourly, daily and weekly basis, achieving accuracy rates above 90 percent and reducing FX costs by as much as 60 percent.

Falcon TST also supports the firm’s liquidity and multi-currency management.

The model has been deployed with partners across sectors, including use cases that help businesses manage FX volatility and assist airlines in keeping fare prices more stable and competitive.

With global air travel expected to reach almost 10 billion passengers in 2025, improved forecasting can help airlines manage cost swings.

Falcon TST can also be applied to weather data, calendar-driven demand shifts, financial market movements and cross-border traffic flows.

Ant International said open-sourcing the model aims to drive wider testing and real-world feedback.

Falcon TST is available on GitHub and Hugging Face, with further documentation on its project website.