WindBorne’s model portfolio includes WeatherMesh-6 Global and WM-6 3km, our next-generation AI models, and MetaMesh, our blended point-forecast model.
Our models
WeatherMesh is our record-breaking AI model. MetaMesh blends WeatherMesh with public AI and NWP models.
AI model
AI model
Blended model
Technical details
WindBorne pairs a global proprietary observation network with an independent AI data assimilation system, giving every model a more complete view of the atmosphere.
State-of-the-art neural network architecture optimized for processing weather data
Encoder-processor-decoder structure enables flexible training and efficient inference
Weather states are mapped into a latent space for efficient pattern recognition
Trained on more than 50 years of historical weather data, augmented by observations from our proprietary network
Verified performance
Earlier signals, more frequent updates, and transparent performance help teams respond as the forecast changes.
WM-6 vs. AIFS ensemble mean
Global forecasts initialized April 2025–October 2026. Snapshot as of October 15, 2026.
Explore the benchmarksForecast accuracy and impact
Proprietary observations, AI-native data assimilation, and high-frequency model cycles are designed to deliver best-in-class forecast accuracy across the full 15-day horizon.
See the benchmarksExtreme events create the greatest operational risk. WindBorne models are designed to surface them earlier. For example, our model flagged Storm Kristin three days before ECMWF and GFS converged.
See case studiesWeatherMesh-6 forecasts update every 15 minutes (WM-6 3km) and hourly (WM-6 Global), replacing discrete six-hour cycles with a continuous forecast signal. This gives teams a continuously updated view as conditions and model confidence evolve.
Delivery
Every forecast product is delivered via REST API and an interactive insights platform. Your automated systems get programmatic access, and your team gets a meteorological-grade interface that shows why a model is producing a given forecast.
from windborne import get_point_forecasts
# By coordinates
point_forecasts = get_point_forecasts(
coordinates="37.77,-122.42")
# By station
point_forecasts = get_point_forecasts(
stations="panc;kjfk")Learn more
WindBorne’s next-generation AI model is the most skillful medium-range weather model to date. Explore the breakthroughs in AI data assimilation and ensemble design behind it.
See WM-6 release notesWeatherMesh-6 saw the Northeast’s strongest blizzard in a decade take shape two days before other models, then sharpened the forecast by the hour.
Read case studyThe only multi-model blend with access to proprietary balloon observations. See how MetaMesh dynamically weights WeatherMesh and operational NWP models.
See MetaMesh blogDuring a major UK windstorm, MetaMesh blended the world’s leading models into one forecast that stayed ahead as conditions evolved.
Read case studyBacktests for both WeatherMesh-6 models are available from April 2025 to present, so you can score on your own variables and regions before going live. Our blended model MetaMesh dynamically recalibrates its weights as new input models arrive, so historical backtests would not accurately represent live model performance. For this reason, MetaMesh is available as a live forecast product only.
WeatherMesh-6 is our AI forecast model, trained on proprietary observations and running hourly. MetaMesh is our blended point forecast product: it combines HRRR, GFS, GFS Ensemble, IFS, IFS Ensemble, AIFS, AIFS Ensemble, and WeatherMesh models with dynamically learned weights that adapt by variable, lead time, region, and season. WeatherMesh is the raw AI signal. MetaMesh is the best-of-breed consensus. It updates on a rolling basis with hourly time steps out to 15 days, optimized for over 5,349 METAR stations with dynamic global coverage.
REST API with JSON and Zarr responses for point and gridded data, plus a Python SDK. Most teams can begin pulling data within a day of onboarding. View API documentation.
We use optional analytics cookies and similar technologies to measure website traffic and understand how visitors use this site.
We’ll remember your choice for 6 months. Learn more
Your browser’s privacy setting keeps analytics off.