Vision Transformers (ViTs) are powerful for image analysis, but high-resolution sensors create massive data challenges. Compression offers a solution, but its impact on ViT performance needs careful evaluation. This presentation explores the performance of ViTs and CNNs on Jetraw-compressed raw image data using modern datasets, models, and simulation. We’ll examine the value of raw images for transformers, Jetraw’s compression principles that preserve performance, and comparative results between ViTs and CNNs on compressed data. We’ll present a corner-case analysis using physically-modeled cameras and challenging conditions. We will demonstrate how intelligent compression maintains accuracy while enabling significant savings in time, cost, and environmental impact, proving its viability for scalable and sustainable computer vision deployments.