Technology / WORLD

Ai2 updates HiRO-ACE for three-kilometre climate simulations

The research team reports lower simulation costs while describing limits to the model’s scope.

Reading sizeSources ↓Share story ↗
Conceptual illustration of technology, processors and connected information.Illustration
Symbolic illustration

Ai2 announced HiRO-ACEv2 on October 14, an update to its system for producing global climate simulations at three-kilometre resolution. The system combines ACE, which models conditions at 100 kilometres, with HiROv2, which downscales those outputs. Ai2 says the updated component handles high-resolution rain, wind and pressure fields and better reproduces extreme events.

In Ai2’s tests against the X-SHiELD simulation used for training, high-resolution statistics were within 3% even for events appearing once per million data points. This result applies to the test in which HiRO received coarsened X-SHiELD data. The institute estimates global runs cost about 1/300 as much as X-SHiELD. Ai2 says this is climate simulation, not operational weather forecasting: the model is tuned to X-SHiELD and cannot yet represent other climates or ERA5 historical data.

Data

Timeline

Announcement summary

The research team reports lower simulation costs while describing limits to the model’s scope.

  1. · Occurred

    Ai2 updates HiRO-ACE for three-kilometre climate simulations

    The research team reports lower simulation costs while describing limits to the model’s scope.

Methodology

This event record summarizes the source announcement and adds no independent verification or generalization beyond it.

Updated:

Structured coverage

Subjects and places

Subject proposals

Geographic references

Only source-provided geographic references are shown. Opening an external map is optional. Optional story maps use OpenStreetMap. Opening a map sends your IP address and this site’s address to the map provider; location data remains readable as text without loading the map.

  • United States · Place role: Coverage · Place type: Country or area · Location precision: Regional
    Evidence
    • Explicit preserved editorial country/region; no city or national centroid inferredallenai.org ↗
Taxonomy version: IPTC Media Topics 2026-Q2

Continue reading

Technology

SlackSwap combines data transfers and recomputation in GPU-limited training

Researchers at UC Merced and the University of Texas at Austin introduced SlackSwap, a planner for deep-learning training. It recomputes some activations when the transfer budget between a GPU and host memory is exceeded; reported results come from simulation, not hardware measurements.

arXiv1 min read
Data included
Technology

AlphaNIST trains a robot assembly policy in simulation

Researchers tested multi-step assembly using one reinforcement-learning policy with sparse rewards.

arXiv preprint by NVIDIA and University of Washington researchers1 min read
Data included