The Problem

Every day, crucial decisions are made based on outdated geospatial simulations.

Insurance companies guess more than they would like about hurricane risks. Airports can't predict tomorrow's delays, let alone next week's. Meanwhile, utilities are crossing their fingers every heatwave that the grid won't collapse.

Earth’s simulation infrastructure needs an upgrade.

The Solution

Today, Silurian is announcing the launch of GFT (Generative Forecasting Transformer), a 1.5B frontier model that simulates global weather up to 14 days ahead at approximately 11km resolution with unprecedented accuracy. Our model outperforms numerical weather prediction systems from the American (NOAA) and European (ECMWF) agencies by up to 30% across multiple variables.

Remember Hurricane Beryl? In a simulation (shown above) run on the 1st of July, GFT correctly predicted that the hurricane would hit Texas. In contrast, the NOAA and ECMWF models wrongly predicted a landfall in Mexico, as The New York Times recently reported.

GFT is immediately available to our customers via our Artificial Planetary Intelligence (API). The API allows users to easily query our models at any location on Earth, and it will become publicly available in the next few weeks.

GFT is only the beginning. We are working towards state-of-the-art simulation infrastructure for verticals such as energy, climate, and agriculture.

Team

We are a team of scientists and engineers with years of experience in building large-scale AI simulations at Microsoft Research and Google Brain. Our work on weather and atmospheric chemistry simulations at Microsoft has been featured in Nature, Bloomberg, and The Wall Street Journal.

Our Ask

Do you want to talk about the weather? Reach out to us at founders@silurian.ai!