More than 40% of endovascular therapy (EVT) fail to achieve complete reperfusion of the territory of the occluded artery in patients with acute ischemic stroke (AIS). Understanding factors influencing EVT could help overcome its limitations. Our objective was to study the impact of thrombus cell composition on EVT procedures, using a simulation system for modeling thrombus-induced large vessel occlusion (LVO) in flow conditions. In an open comparative trial, we analyzed the behavior of size-standardized platelet-rich and red blood cells (RBC)-rich thrombi during simulated stent retriever-mediated EVT procedures. Sixteen simulated EVT procedures were performed (8 RBC- vs. 8 platelet-rich thrombi). Platelet-rich thrombi were associated with a higher number of stent retriever passes (p = 0.03) and a longer procedure duration (p = 0.02) compared to RBC-rich thrombi. Conversely, RBC-rich thrombi released more embolic fragments than platelet-rich thrombi (p = 0.004). Both RBC-rich and platelet-rich thrombi underwent drastic compaction after being injected into the in vitro circulation model, and histologic analyses showed that these EVT-retrieved thrombi displayed features comparable to those previously observed in thrombi from patients with AIS patients having LVO, including a marked structural dichotomy between RBC- and platelet-rich areas. Our results show that the injection of in vitro-produced thrombi in artificial cerebrovascular arterial networks is suitable for testing recanalization efficacy and the risk of embolization of EVT devices and strategies in association with thrombus cell composition.