WeatherNext 3 is Google DeepMind and Google Research's global AI weather forecasting model. It produces hourly forecasts at up to 5 km resolution using live satellite data.
Web appContact for pricing
Step 1: Open the WeatherNext developer page and read the model specifications.
Step 2: Use the quick start guide to access forecast data in Cloud Storage, Earth Engine or BigQuery.
Step 3: Run custom inference on Google Cloud if you need your own ensemble size or forecast horizon.
Step 4: Explore the open source research models and notebooks on GitHub.
Step 5: Try Weather Lab to view live forecasts and cyclone tracks.
WeatherNext 3 generates a new forecast every hour using live geostationary satellite data.
Temperature and moisture are at 5 km, other surface variables at 10 km and atmospheric variables such as wind at 25 km.
Forecast data is offered through Google Cloud Storage (Zarr), Earth Engine and BigQuery. Custom inference runs on Google Cloud.
Google says it is integrated across Search, Gemini, Maps, Google Maps Platform and Cloud.
Almanac MCPAlmanac computes magnetic declination and sky positions deterministically and exposes them to AI agents as an MCP server. Results are checked against the publishing authorities.
OrchestKit Docs MCPA remote MCP server for the OrchestKit documentation, with read-only tools to search the docs and fetch pages as Markdown. OrchestKit is a free, open-source Claude Code plugin.