Floods are catastrophic events that affect nearly every part of the globe, leading to substantial losses of life and causing extensive economic damage. During flood emergencies, one of the primary challenges faced by response teams is accurately identifying the flooded regions to establish access points and determine safe evacuation routes swiftly. As climate change intensifies, the frequency and severity of extreme weather events like floods are rising, making rapid flood mapping and response even more critical in minimizing damage and saving lives. This study compares several artificial intelligence models and spectral indexes for flood detection in Sentinel 2 multispectral images. The proposed methodology combines transfer learning with different models, such as neural networks, convolutional neural networks, and vision transformers. A variation of this model is the vision transformer (ViT), which can be applied to image classification tasks. Multispectral Instrument (MSI) images from Sentinel-2 contain images from different bands. By combining different zones of those bands different spectral indexes can be calculated. This study uses the Normalized Difference Water Index (NDWI), Modified Normalized Difference Water Index (MNDWI), Water Ratio Index (WRI), Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Automated Water Extraction Index (AWEI). By comparing different artificial intelligence models and different spectral indexes the best combination is determined, which is NDVI with VGG CNN model, and can be used for real-time flood detection using Sentinel 2 Multispectral Images.