Back    Zoom +    Zoom -
Ant Int'l Opens Source of 'Falcon' AI Predictive Model with 90%+ Accuracy
Recommend
19
Positive
39
Negative
9
Ant International announced the official open- source of its independently developed "Falcon" AI predictive model, which is the industry's first large-scale time series predictive foundational model based on a multi-segment (Patch) approach and utilizing a "Mixture of Experts" (MoE) architecture, with a parameter scale exceeding 2.5 billion.

Currently, the "Falcon" AI predictive model is fully available on GitHub, Hugging Face, and Ant International's platform for global developers and research institutions to use.

Related NewsCICC Lists H-shr Top Picks for 2026 (Table)
The "Falcon" AI predictive model was initially used internally by Ant International for cash flow and FX risk predictions. The model can make predictions on an hourly, daily, or weekly basis with an accuracy rate exceeding 90%, helping companies reduce FX costs by up to 60%.

In addition to the financial sector, the "Falcon" AI predictive model can also be used to predict weather changes, holiday consumption, financial market fluctuations, cross-border human flow, and other data with time series characteristics.

The company is now exploring specific applications with partners in industries such as aviation, banking, online travel platforms, and e-commerce.

Related NewsHaitong Int'l Expects HK Stocks to Reach New Heights; Focus Mainly on Techs w/ Attention to Innovative Drugs & Brokers

AASTOCKS Financial News
Website: www.aastocks.com