Accurate characterization of individual fire events is critical for understanding global fire regimes and their interactions with climate. However, existing global inventories relying on coarse-resolution data and annual-aggregation algorithm may obscure small-scale thermal anomalies and artificially fragment prolonged fires. Here, we present a global dataset of individual fire events for 2012–2025 derived from Suomi-NPP VIIRS 375 m active fire detections. To address the spatiotemporal discontinuities in conventional clustering approaches, we develop a tracking framework based on a sliding-window method that merges spatially overlapped and temporally continuous fire patches. This approach maintains the continuity of long-duration events and reduces artificial segmentation, as demonstrated by validation against ground-based and higher-resolution satellite benchmarks. Our dataset provides a range of event-level metrics, including ignition location, fire duration, daily expansion rates, burned area, and fire radiative power. This fire event inventory offers an observational foundation for tracking extreme fire behaviour, calibrating fire spread models, and refining global emissions estimates.