Despite the diverse nature of clinical trajectories after a first-episode of psychosis, few baseline characteristics have been predictive of clinical improvement, and the neurobiological underpinnings of this heterogeneity remain largely unknown. Elevated extracellular free water (FW) in the brain is a diffusion imaging measure that has been consistently reported in different phases of psychosis that may indicate a neuroinflammatory state. Its predictive capacity in terms of clinical outcomes, however, is unknown. We used diffusion imaging to determine FW and tissue-specific fractional anisotropy (FA-t) in first-episode psychosis. Forty-seven participants were categorized as clinical "Improvers" (n=26) if they achieved a 20% decrease in total Brief Psychiatric Rating Scale (BPRS) score at 12 months. To determine the predictive capacity of FW and FA-t, these measures were introduced in a stepwise logistic regression model to predict clinical improvement. For those measures surviving the model, regional between-group differences were also investigated in cortical surface or white matter tracts as applicable. Higher gray matter (GM) FW (OR-CI 1.134 - 2.543) and FA-t (OR-CI: 0.905 - 2.038) both predicted Improver status. FW in GM also linearly correlated with BPRS total score at 12 months follow-up. Examining regional specificity, Improvers showed greater FW predominantly in temporal regions and higher FA-t values in several white matter tracts, including bilateral longitudinal superior fasciculus. Our results show that elevated FW in GM and FA-t predict further clinical improvement during the initial phases of psychosis. The potential roles of brain inflammatory processes in predicting clinical improvement are discussed.