Ant International Open Sources Falcon TST AI Forecasting Model
Ant International has officially open-sourced its proprietary Falcon TST (Time-Series Transformer) AI model, marking a significant development in the field of artificial intelligence and big data analytics. This industry-first model utilizes a Mixture of Experts architecture with multiple patch tokenizers and is supported by up to 2.5 billion parameters. Initially deployed internally to manage cashflow and foreign exchange exposure, the model has demonstrated over 90% accuracy and reduced FX costs by up to 60%. It also achieves state-of-the-art zero-shot results on long-term forecasting benchmarks. Jiang-Ming Yang, Chief Innovation Officer at Ant International, emphasized that releasing the model on GitHub and Hugging Face aims to foster global collaboration among scientists and developers. Beyond financial applications, Falcon TST supports diverse time-series forecasts, including weather patterns, calendar events, and cross-border traffic data. Industry partners are already leveraging the technology to mitigate volatility and help airlines offer stable pricing, potentially benefiting the projected 10 billion air travelers in 2025. This move underscores Ant International's commitment to advancing AI utility through open-source community contributions.
Wire timeline
Ant International Open Sources Falcon TST AI Forecasting Model
Ant International has officially open-sourced its proprietary Falcon TST (Time-Series Transformer) AI model, marking a significant development in the field of artificial intelligence and big data analytics. This industry-first model utilizes a Mixture of Experts architecture with multiple patch tokenizers and is supported by up to 2.5 billion parameters. Initially deployed internally to manage cashflow and foreign exchange exposure, the model has demonstrated over 90% accuracy and reduced FX costs by up to 60%. It also achieves state-of-the-art zero-shot results on long-term forecasting benchmarks. Jiang-Ming Yang, Chief Innovation Officer at Ant International, emphasized that releasing the model on GitHub and Hugging Face aims to foster global collaboration among scientists and developers. Beyond financial applications, Falcon TST supports diverse time-series forecasts, including weather patterns, calendar events, and cross-border traffic data. Industry partners are already leveraging the technology to mitigate volatility and help airlines offer stable pricing, potentially benefiting the projected 10 billion air travelers in 2025. This move underscores Ant International's commitment to advancing AI utility through open-source community contributions.
TechNode