FLASH Weather AI has launched a new high-resolution hail prediction model that can give almost an hour of warning before a storm.

FLASH’s 1-Hour Hail Swath Prediction Model is a deep-learning system that forecasts hail size and arrival time at 1-kilometer resolution, refreshed every five minutes, with predictions extending up to 55 minutes into the future.

Trained on four years of U.S. convective storm data spanning 2021 through 2024, the model produces deterministic forecasts of hail size in inches or millimeters, predicting down to the hundredth of an inch. A companion Hail Arrival Time product estimates when quarter-size or larger hail will reach any location in the forecast window.

FLASH said that after a month of beta testing across multiple severe weather events, the model has demonstrated significant accuracy and operational usability, with verification examples available on request.

The product began rolling out this week with API access as well as inside the FLASH Weather Command Center (WCC), the company’s web-based operations platform, with native integration into the FLASH mobile app following in the coming weeks.

“For the first time, insurers, fleet operators, and infrastructure managers have a tool that doesn’t just tell you a storm is coming; it tells you exactly where hail will fall, how large it will be, and gives you enough lead time to act,” said Jason Deese, founder & CEO of FLASH Weather AI, in a statement.

“That’s not an incremental improvement,” he said. “That’s a fundamental shift in how the world manages hail risk.”

FLASH said the launch establishes the foundation for a full hail prediction suite. Upcoming features include hail probability forecasts that mirror the company’s existing lightning probability product, push notifications for hail threats inside the FLASH mobile app, and ongoing model retraining as additional convective season data becomes available.