Generative climate model compresses a century into a day of GPU time.
UC San Diego and Allen Institute for AI released Spherical DYffusion, a diffusion model fine-tuned on ERA5 reanalysis data. The model ingests 3-D spherical harmonics and emits 100-year temperature, precipitation, and wind fields in 25 GPU-hours on 8 A100s. Physics residuals stayed under 0.8 percent against CMIP6 ensemble means.
You treat long-horizon simulation as a forward pass rather than a months-long HPC queue. Your workflow now includes a quick generative pre-check before committing to full physics runs.
Allen Institute climatologists used the model to produce 1,000-member ensemble members for the 2025 State of the Climate report; the ensemble finished in four days instead of the previous six-week schedule.
Step 1: Clone the repo at github.com/allenai/spherical-dyffusion. Step 2: Edit config.yaml to set horizon=100 and ensemble_size=50. Step 3: Execute python run_inference.py --gpus 4; the script writes NetCDF output and prints wall-clock time under 30 minutes on a single node.