A few tau immunotherapies are now in clinical trials with several more likely to be initiated in the near future. A priori, it can be anticipated that an antibody which broadly recognizes various pathological tau aggregates with high affinity would have the ideal therapeutic properties. Tau antibodies 4E6 and 6B2, raised against the same epitope region but of varying specificity and affinity, were tested for acutely improving cognition and reducing tau pathology in transgenic tauopathy mice and neuronal cultures. Surprisingly, this study shows that one antibody, 4E6, which has low affinity for most forms of tau acutely improved cognition and reduced soluble phospho-tau, whereas another antibody, 6B2, which has high affinity for various tau species was ineffective. Concurrently, they confirmed and clarified these efficacy differences in an ex vivo model of tauopathy. Alzheimer’s paired helical filaments (PHF) were toxic to the neurons and increased tau levels in remaining neurons. Both toxicity and tau seeding were prevented by 4E6 but not by 6B2. Furthermore, 4E6 reduced PHF spreading between neurons. Interestingly, 4E6’s efficacy relates to its high affinity binding to solubilized PHF, whereas the ineffective 6B2 binds mainly to aggregated PHF. Blocking 4E6's uptake into neurons prevented its protective effects if the antibody was administered after PHF had been internalized. When 4E6 and PHF were administered at the same time, the antibody was protective extracellularly. Overall, these findings indicate that high antibody affinity for solubilized PHF predicts efficacy, and that acute antibody-mediated improvement in cognition relates to clearance of soluble phospho-tau. Importantly, both intra- and extracellular clearance pathways are in play. Together, these results have major implications for understanding the pathogenesis of tauopathies and for development of immunotherapies.