Because many serious adolescent offenders reduce their antisocial behavior after court involvement, understanding the patterns and mechanisms of the process of desistance from criminal activity is essential for developing effective interventions and legal policy. This study examined patterns of self-reported antisocial behavior over a 3-year period after court involvement in a sample of 1,119 serious male adolescent offenders. Using growth mixture models, and incorporating time at risk for offending in the community, we identified five trajectory groups, including a "persister" group (8.7% of the sample) and a "desister" group (14.6% of the sample). Case characteristics (age, ethnicity, antisocial history, deviant peers, a criminal father, substance use, psychosocial maturity) differentiated the five trajectory groups well, but did not effectively differentiate the persisting from desisting group. We show that even the most serious adolescent offenders report relatively low levels of antisocial activity after court involvement, but that distinguishing effectively between high-frequency offenders who desist and those who persist requires further consideration of potentially important dynamic factors related to this process.