Lithium-induced neurotoxicity may be life threatening. Three patterns have been described, including acute, acute-on-chronic, and chronic poisoning, with unexplained discrepancies in the relationship between clinical features and plasma lithium concentrations. Our objective was to investigate differences in plasma, erythrocyte, cerebrospinal fluid, and brain lithium pharmacokinetics using a multicompartmental approach in rat models mimicking the three human intoxication patterns. We developed acute (intraperitoneal administration of 185 mg/kg Li2CO3 in naive rats), acute-on-chronic (intraperitoneal administration of 185 mg/kg Li2CO3 in rats receiving 800 mg/l Li2CO3 in water during 28 days), and chronic poisoning models (intraperitoneal administration of 74 mg/kg Li2CO3 during 5 days in rats with 15 mg/kg K2Cr2O7-induced renal failure). Delayed absorption (4.03 vs 0.31 h), increased plasma elimination (0.65 vs 0.37 l/kg/h) and shorter half-life (1.75 vs 2.68 h) were observed in acute-on-chronically compared with acutely poisoned rats. Erythrocyte and cerebrospinal fluid kinetics paralleled plasma kinetics in both models. Brain lithium distribution was rapid (as early as 15 min), inhomogeneous and with delayed elimination (over 78 h). However, brain lithium accumulation was more marked in acute-on-chronically than acutely poisoned rats [area-under-the-curve of brain concentrations (379 ± 41 vs 295 ± 26, P < .05) and brain-to-plasma ratio (45 ± 10 vs 8 ± 2, P < .0001) at 54 h]. Moreover, brain lithium distribution was increased in chronically compared with acute-on-chronically poisoned rats (brain-to-plasma ratio: 9 ± 1 vs 3 ± 0, P < .01). In conclusion, prolonged rat exposure results in brain lithium accumulation, which is more marked in the presence of renal failure. Our data suggest that differences in plasma and brain kinetics may at least partially explain the observed variability between human intoxication patterns.