Organ and cancer segmentation in medical images, especially from 3D CT and MR scans, is fundamentally important for accurate diagnosis, treatment planning, and monitoring the progression of diseases. Precise segmentation results of organs and pathological lesions can aid clinicians in formulating personalized treatment strategies, which are essential for optimal patient outcomes. Deep learning-based methods have significantly revolutionized these tasks, achieving unprecedented levels of accuracy and automation compared to traditional model-based methods. However, a notable limitation is that most deep learning models are tailored for specific types of cancer, such as brain cancer, lung cancer, liver cancer, and so on. As a result, the generalizability of these algorithms across various cancer types remains a challenge. Another main barrier that hinders the real-world deployment of existing methods is the algorithm's efficiency because deep learning models usually require considerable computing for running, such as GPU, CPU, and RAM. During the past three years, we have organized three challenges to address these limitations with community efforts. FLARE 2021: four abdominal organs segmentation in CT scans; Data: 511 CT scans FLARE 2022: 13 abdominal organs segmentation in CT scans; Data: 2300 CT scans FLARE 2023: 13 abdominal organs and pan-cancer segmentation in CT scans; Data: 4500 CT scans Nowadays, the winning solution can simultaneously segment 13 organs and various abdominal lesions within 10 seconds for a 3D CT scan with over 1,000,000 voxels, which significantly improved the segmentation accuracy, efficiency, and generalization ability. In FLARE 2024, we aim to further promote the development of pan-cancer segmentation and model deployment on low-resource settings (e.g., no GPU available) by extending the challenge to the following three tasks: Subtask 1: Pan-cancer segmentation in CT scans Subtask 2: Abdominal CT organ segmentation on laptop Subtask 3: Unsupervised domain adaptation for abdominal organ segmentation in MRI Scans We will provide a comprehensive and large-scale dataset with more than 10,000 CT scans and 1,000 abdominal MR scans, which is a multi-racial, multi-center, multi-disease, multi-phase, multi-manufacturer, and multi-modality dataset. This challenge would mark a significant stride for universal cancer segmentation models and applicable toolsets for CT and MR image analysis.