Timestral

TimesFM-3 is a state-of-the-art time series foundation model designed for highly accurate multivariate forecasting in a single forward pass. The 330 million parameter model is pre-trained on a real-world and synthetic time-series corpus comprising more than 1 trillion time points, building on the efficiency and zero-shot generalization of earlier TimesFM models. It can jointly predict multiple coevolving time series and capture dependencies that improve accuracy without task-specific fine-tuning. The model supports multiple targets with point and quantile forecasts, past covariates that are known only historically, and past-future dynamic covariates such as planned promotions, holidays, or weather forecasts. TimesFM-3 uses a decoder-only transformer architecture, processes contiguous data in patches of 32 time steps, and applies alternating causal temporal attention and full variate attention to combine patterns across time and related series.

NVIDIA Earth-2 is a high-performance, AI-accelerated global weather and climate forecasting platform that combines open models, libraries and frameworks to make professional-grade forecasting accessible and deployable on modern infrastructure instead of requiring traditional supercomputers. It provides end-to-end capabilities from ingesting raw observational data to producing high-resolution weather and climate predictions, including local storm forecasts and up to 15-day medium-range forecasts, with generative AI architectures that drastically speed computations while maintaining or improving accuracy compared with conventional methods. The Earth-2 stack consists of models like Medium Range (Atlas architecture for multi-variable forecasts), Nowcasting (StormScope for 0-to-6-hour local weather), Global Data Assimilation (HealDA for initial atmospheric states), CorrDiff (AI downscaling to regional resolution), and FourCastNet 3 (fast global forecasting).