👋 Hello, team! I am doing some tests with timegpt and I was wondering at what forecast horizons should I consider switching from timegpt-1 to timegpt-1-long-horizon for hourly, daily, weekly, monthly, and yearly data? Are there specific thresholds or guidelines for each time frequency? I could not find this on the documentation

Hello! There is no hard rule for that, but when you forecast more than two seasonal periods, it can be considered as a long horizon, so you should switch model. For example, if you have hourly data with a daily seasonlity, then forecasting on a horizon longer than 2 days is considered "long horizon". I hope this helps!

Got it, thanks for the response @Marco!