Ant International Unveils Major Upgrade to Time-Series AI Model, FalconTST 2.0 Sets New Global Benchmark

A Leap Forward in Time-Series Forecasting: Ant International Launches FalconTST 2.0

Ant International has officially released version 2.0 of its self-developed large time-series prediction model, FalconTST. This upgrade represents a significant refinement, specifically engineered to tackle complex business challenges, with a strong initial focus on the cross-border finance sector.

Beyond Cross-Border Finance: Expanding the Model's Reach

While the model's primary design tackles foreign exchange risk management for cross-border transactions—enabling institutions to better anticipate currency fluctuations—its applications are growing. Plans are underway to extend this technology to other critical areas such as demand forecasting in e-commerce supply chains and operational management for airlines. This expansion highlights the model's robust capability to handle the common complexities of time-series data across diverse industries.

Setting a New Global Benchmark: Unmatched Performance

A core metric for evaluating time-series models is the Mean Absolute Scaled Error (MASE), where a lower score indicates greater prediction accuracy. In recent global benchmark tests, FalconTST 2.0 achieved a state-of-the-art (SOTA) result, lowering the MASE to 0.666. This performance surpasses comparable models from several leading global tech companies, cementing its technical superiority.

Validated by the Market: Adoption by Leading Financial Institutions

The true test of any technology is its practical adoption. Several major international financial institutions, including Barclays, Citi, Deutsche Bank, and Standard Chartered, are now implementing FalconTST 2.0. They are primarily applying it to core financial operations like corporate cash flow forecasting and dynamic foreign exchange risk management, leveraging AI to enhance the foresight and precision of their decisions.

From technical breakthrough to real-world implementation, the launch of FalconTST 2.0 signals AI's rapid move from the lab to the heart of industry in the specialized field of time-series data analysis.