Extreme wildfire events (EWEs), defined here as wildfires exceeding 5,000 ha, are a growing concern across European landscapes, posing important threats to economies, ecosystems and human safety. Accurate and timely warnings are essential to prevent and mitigate their devastating impacts. We leveraged advanced probabilistic modelling to forecast the daily occurrence of EWEs at an 8 km resolution across Europe up to 9 days in advance. Our model significantly improved the estimation of EWE probability while providing key insights into the environmental and climatic drivers of EWEs. Compared to the Fire Weather Index (FWI), our model increased precision by a factor 5 for EWE predictions at the NUTS3 level and by a factor of 10 at 8km resolution”. Precision, defined as the proportion of correct EWE predictions among all EWE alerts, was evaluated at a detection threshold that captures 50% of observed events. To assess its operational relevance, we evaluated the model’s forecasting skill in a test area. Our results demonstrate that EWE forecasts up to nine days in advance outperformed FWI-based nowcasts, emphasizing the model’s potential for operational use and its superiority in anticipating EWEs compared to current fire danger rating systems.