Weather Lab is the interactive research platform Google published alongside the WeatherNext 3 launch: a place to explore the model's global forecast layers and compare them with meteorological baselines while evaluations continue. It is the closest view of how the model behaves as a system - and it is a research window rather than a weather product. This page explains what it is for, how to use it well and where its limits are. For the launch context, see the release timeline; for the science behind the layers, see the paper's key points.
7-day free trial, then $19/mo · cancel anytime · compare plans

The platform is built around one idea: look at the model directly, and look at it critically.
The model's worldwide fields are exposed for exploration - built for seeing how the forecast behaves across the planet, not just where you live.
Layers can be compared against meteorological baselines, which is how you spot where the model adds skill and where it still does not.
It sits alongside the ongoing evaluation process - the same model runs that independent leaderboards score as observations arrive.
Built for exploration rather than alerts, accounts or exports: closer to a lab bench than to a weather app.
A few habits separate a useful session from a misleading one.
Pick a single layer - temperature, precipitation or wind - and follow it across a week before stacking more variables on top of each other.
Compare the model's picture for your own area against what you remember actually happened there. Familiar ground is the best calibration.
Use the comparison layers continuously. The model's story reads differently when a physical baseline disagrees at the same moment.
The layers reflect a model under continuing evaluation. Note the date on anything you cite, because the picture can change as the system updates.
Clearing up what the platform deliberately does not try to be.
Official alerts come from meteorological agencies. Weather Lab will not page anyone when a storm turns, and it is not built to.
No plans, credits, saved locations or exports. Those needs belong to forecast products, including independent workspaces like this one.
Research views can lag or differ from what operational surfaces serve. When freshness matters, check the timing before acting on a layer.
It is a window into model behavior, not a verification suite. The paper and the live leaderboards remain the reference for skill claims.
Short answers about the research platform and how it relates to forecasts.
Next: the paper's key points explain the science behind the layers.
Weather Lab shows how the model sees the planet; an hourly workspace shows what that means for your coordinates. Start the 7-day free trial, or see how to read an hourly forecast first.