Ant International, a leading global digital payment, digitization, and financial technology provider, has open-sourced its proprietary Falcon TST (Time-Series Transformer) AI model, marking a significant step toward democratizing advanced forecasting capabilities across industries. The model, built on a novel Mixture of Experts architecture with multiple patch tokenizers and supporting up to 2.5 billion parameters, delivers state-of-the-art zero-shot performance on major long-term forecasting benchmarks, including achieving record-low absolute mean error rates. Internally deployed at Ant International, Falcon TST powers critical financial forecasting tasks such as cashflow prediction and foreign exchange (FX) exposure management, delivering accuracy exceeding 90% across hourly, daily, and weekly forecasts. As a result, the company has reduced its FX-related costs by up to 60%. Beyond internal use, Ant International has partnered with various industry players to apply the model in real-world scenarios. These include helping businesses stabilize FX costs and manage volatility, as well as enabling airlines to offer more predictable and competitive pricing. With global air travel projected to reach nearly 10 billion passengers by 2025—according to the Airports Council International World—these AI-driven optimizations could lead to substantial savings for consumers worldwide. The Falcon TST model is designed to support a broad range of time-series forecasting applications, including weather patterns, calendar events, financial market movements, and cross-border traffic data, making it a versatile tool for enterprises across sectors. “We believe the future of AI lies in collaboration,” said Jiang-Ming Yang, Chief Innovation Officer at Ant International. “By open-sourcing Falcon TST, we’re inviting scientists, developers, and researchers from around the world to contribute, test, and improve the model, accelerating innovation in time series learning and expanding its impact.” The Falcon TST AI Model is now available on GitHub and Hugging Face, where developers can access the code, documentation, and training resources. Ant International encourages the global tech community to explore the model and share insights through real-world applications. With headquarters in Singapore and operations spanning Asia, Europe, the Middle East, and Latin America, Ant International continues to drive inclusive growth by empowering financial institutions and merchants of all sizes with cutting-edge digital payment and financial technology solutions. For more information, visit https://www.ant-intl.com/.