Background: There are many multiple myeloma (MM) treatments and nearly all patients undergo continual cycles of treatment, response, and resistance. Selecting an effective therapeutic strategy is of critical importance; but there is currently a lack of patient-derived MM models that can enable functional precision medicine to help real-time clinical decision-making to guide individual patient treatment. Methods: We show a method to rapidly establish MM patient-derived MicroOrganoSheres (MOS), which are microscale, droplet-sized patient avatars that sustain the native tumor niche, including both stromal and immune compartments. We generated first-of-its-kind data showing that MM MOS predicted patient treatment outcomes and have the potential to guide clinical treatment decisions. We further developed high-throughput drug screen to demonstrate MM MOS response to FDA-approved single and combination therapies and developed MOS assays to predict patient responses to immunotherapies including bispecific and phagocytosis-inducing antibodies. Results: MM MOS established from patient bone marrow (BM) biopsies preserved CD138/CD38 tumor cells, stromal (e.g., osteoblast) cells, and all major immune cell populations (T, NK, B, macrophage, dendritic, and myeloid-derived suppressor cells) for 11 days in culture. We performed single and combo drug testing on MOS derived from 7 MM patients within 10 days of biopsy collection. Patients went on receiving treatments and their responses were assessed using International Myeloma Working Group response criteria. The MOS assay predicted responses of both treatment-naïve and refractory patient to standard-of-care regimens, including combinations of proteasome inhibitors (bortezomib, carfilzomib), immunomodulatory agents (lenalidomide) and DCEP. Moreover, the BM biopsy-derived MOS enabled high-throughput drug testing of single agents and combinations in 9-dose titrations, providing clinicians the ability to evaluate alternative treatment options. Furthermore, because MOS retain the immune compartments of the original tumor niche, we were able to evaluate patients’ response to bispecific T cell engagers and developed a phagocytosis assay to measure anti-tumor phagocytosis activity, which showed efficacy in both single-agent and combo settings. Multi-modal profiling and an AI image analysis pipeline provided robust measurements that confirmed the unique phenotypic profile of each patient sample, highlighting the critical unmet need to guide each patient to the optimal treatment. Conclusions: Based on this groundbreaking feasibility study, a subsequent clinical trial of 40 patients is current under IRB review to further validate the clinical power of MM MOS to predict patient outcome and to demonstrate utility as a viable functional precision medicine approach to inform treatment decisions in the clinic.