Ai2 Introduces HiRO-ACEv2 Weather System

HuggingFace BlogThursday, October 15, 2026
  • •Ai2 introduced HiRO-ACEv2 to generate global climate data at 3 km resolution for about 1/300th of X-SHiELD’s cost.
  • •HiROv2 matches high-resolution statistics within 3%, including events occurring once in a million data points.
  • •Ai2 plans training for a 4°C warmer world and additional temperature, snow, and ice outputs.

Ai2 introduced HiRO-ACEv2 on October 14, 2026, a system for generating global climate data at 3 km resolution for about 1/300th the cost of X-SHiELD, a high-resolution atmospheric model that explicitly simulates storms and other small-scale weather. HiRO-ACE, introduced earlier in 2026, pairs ACE, which simulates at 100 km resolution, with HiRO, which refines those outputs to 3 km. The update’s HiROv2 predicts several high-resolution weather fields at once instead of rainfall alone and improves predictions of extreme events.

Ai2 says HiROv2 reproduces high-resolution statistics within 3% of coarsened X-SHiELD data, including events that occur only once in a million data points. It estimates tropical cyclones’ total energy, including storms absent from its training data, and generates hurricanes with realistic eyes and rain bands; winds and pressure around the eye remain more lopsided than in X-SHiELD. It also produces strong winds flowing down mountain slopes. In a 10-year simulation, updated ACE’s hurricanes, average climate patterns, and rain, wind, and pressure distributions align well with X-SHiELD.

Together, ACE and HiROv2 generated a full year of global 3 km data whose statistics closely match X-SHiELD, although HiRO used ACE’s own simulation rather than X-SHiELD data. Across a 10-year run, hurricane energy was comparable to X-SHiELD’s, and snapshots showed other similar extreme events. Ai2 says regional runs can cost more than 100 times less, depending on area size, and it is working on a streamlined HiROv2 that could be more than 10 times faster. The system is limited to the ACE version tuned to X-SHiELD, so it cannot yet produce data for other climates, including a warmer future or ERA5’s historical record. Ai2 plans to train on additional X-SHiELD datasets, including one for a world 4°C warmer, and add near-surface temperature, snow, and ice outputs. A web demo compares pre-generated HiROv1 and HiROv2 precipitation, wind, and sea-level pressure with X-SHiELD; models and tutorial notebooks are available to download.