WindBorne Systems, a startup founded by Stanford students in 2019, released its WeatherMesh 6 AI weather forecasting model on June 1, 2026, claiming it outperforms the systems developed by the European Centre for Medium-Range Weather Forecasting (ECMWF), the organization widely regarded as the global leader in weather prediction. The company says WeatherMesh 6 produces forecasts every hour at a resolution of 3 kilometers in Europe and the continental United States, compared to the six-hour update cycle of traditional models. According to WindBorne's Chief Product Officer Kai Marshland, the model "is as accurate five days out as a traditional forecast is the day before," particularly when measuring surface temperatures. The startup originally focused on building improved weather balloons to collect atmospheric data for sale. However, after AI-powered weather forecasting models emerged in 2022, WindBorne shifted strategy to develop its own machine learning system that could process its unique data assets more effectively. Traditional weather forecasting relies on complex physics simulations requiring expensive supercomputers, while AI models like WeatherMesh operate faster but have historically struggled with resolution and predicting distant time horizons accurately. Other major players including Google DeepMind are also developing weather forecasting AI systems, and the technology is already being integrated into government agencies worldwide. WindBorne differentiates itself by combining proprietary data collection with its own modeling capabilities, creating what the company describes as a more comprehensive approach than competitors relying solely on publicly available weather information.