Ai2 says it will build an open climate model linking air and ocean
Ai2 says a three-year project, with funding from Google.org and technical support from Google Research, will extend an AI climate model to the ocean. GeekWire reports the hardest part could take longer than three years.
A research team says it will build a climate model that simulates the atmosphere and the ocean together, aiming for faster projections over years to decades.
Ai2 (the Allen Institute for AI), which GeekWire describes as based in Seattle, says in a post dated 4 November 2026 that it is starting a new climate modelling project. It has three-year funding from Google.org and technical support from Google Research.
Ai2 says the model, called NeuralGCM+, will be open source. It says it plans to release its models and modelling frameworks with documented workflows, so outside researchers can train and adapt models and contribute code.
Ai2 says conventional climate models need so much computing that it limits the experiments scientists can run, and that faster, less expensive simulations could let researchers explore more possible futures.
NeuralGCM+ builds on Google Research's NeuralGCM, which uses physics equations to simulate large-scale movements of the atmosphere, and AI for smaller processes such as cloud formation and rainfall. Ai2 says the AI part looks only at local processes inside vertical slices of the atmosphere. This gives it a simpler problem to learn and may help it cope as the climate warms.
Ai2 says it will extend the model to the ocean, to capture how oceans absorb heat and move it around. Its stated focus is climate trends over years to decades, including how rainfall and water availability could change through mid-century in regions such as the southwestern United States.
The next two figures come from existing models, not NeuralGCM+. Ai2 says that in a public benchmark, a shared test, an atmosphere-only version of NeuralGCM simulated a year of atmospheric conditions for the whole globe in 8 minutes on a single TPU, a computer chip used for AI.
Ai2 also says that in one study its AI climate model, called ACE2, completed experiments on how ocean-temperature patterns affect climate roughly 100 times faster than the conventional atmospheric model it was trained to copy.
The first step, GeekWire reports, is releasing the code for training NeuralGCM, which has been tied to Google's internal systems. Ai2 expects to release that code within the first year, with ocean-model work running at the same time.
GeekWire reports that making the model reliable in a warmer world is the hardest part, and that it could take longer than three years. Ai2 says models trained only on past data struggle to make reliable projections of future climates that have not yet happened, and that demonstrating reliable future projections remains a central research task.
Oliver Watt-Meyer, Climate Modeling Lead at Ai2, said in GeekWire's report: “We don’t know if it’ll work necessarily.”
Sources
- Google climate partnership · allenai.org
- Ai2 gets 5m from google to turn ai weather model into a tool for projecting… · geekwire.com