Allen AI has unveiled HiRO-ACEv2, an AI-powered climate modeling system that can generate high-resolution extreme weather simulations at a fraction of traditional computational costs. The updated system, which combines a coarse-resolution climate model with an AI downscaling model, produces global weather data at 3-kilometer resolution for approximately 1/300th the cost of X-SHiELD, a conventional high-resolution atmospheric simulation model. The breakthrough addresses a major bottleneck in climate research, where detailed weather modeling typically requires prohibitive computational resources. The system consists of two components: ACE, which operates at 100-kilometer resolution, and HiRO, an AI model that sharpens those outputs into 3-kilometer detail. The upgraded HiROv2 can now predict multiple weather fields simultaneously—including rainfall, wind speed, and atmospheric pressure—rather than focusing on a single parameter. Critically, the model excels at reproducing rare extreme weather events, matching high-resolution statistics within 3 percent accuracy even for phenomena that occur only once in millions of data points. In testing over a 10-year simulation period, HiRO-ACEv2 accurately generated hurricane and tropical cyclone patterns, including realistic storm eye structures and wind patterns around mountainous terrain. By dramatically reducing the computational cost of high-resolution climate modeling, HiRO-ACEv2 could democratize access to detailed climate simulation capabilities for research institutions with limited budgets. The system is available as open-source models with tutorial notebooks, alongside a web demo allowing researchers to explore output data. Allen AI indicated future work will expand the system's capabilities to handle different climate scenarios, including simulations of a warmer planet and historical climate datasets, as well as adding outputs for near-surface temperature and snow-ice modeling.