Purpose: This study aimed to investigate the efficacy and tolerability of aripiprazole, olanzapine and risperidone in first-episode schizophrenia (FES). Methods: The eight-week, open, randomised study was conducted in six Chinese medical centres. Altogether, 498 FES subjects were randomised to aripiprazole (n = 165), olanzapine (n = 168) or risperidone (n = 165). Efficacy was measured with the Positive and Negative Syndrome Scale (PANSS), tolerability with the Udvalg for Kliniske Undersogelser Side Effect Rating Scale (UKU) and functioning with the Personal and Social Performance Scale (PSP). Results: All three antipsychotics significantly improved the baseline to end-point PANSS total and each of the sub-scale scores (p < 0.001). Risperidone was superior to olanzapine and aripiprazole regarding PANSS total end-point scores (p < 0.05). Cumulative response (PANSS total score reduction ⩾30%) was similar between risperidone, olanzapine and aripiprazole (74.8%, 73.5% and 70.1%; p = 0.707), but risperidone was superior to aripiprazole regarding PANSS total score reduction ⩾50% (37.8% vs. 26.6%; p < 0.05). Olanzapine was associated with the largest weight gain at week 4 and 8 (p < 0.01), weight gain ⩾7% (olanzapine = 49.0% vs. risperidone = 32.5% vs. aripiprazole = 17.0%; p < 0.01), more psychic side effects at week 8 (p < 0.01 each) and more ‘other’ side effects at week 4 (p < 0.001) and week 8 (p < 0.05) but fewer neurological side effects at week 4 (p < 0.05) and week 8 (p < 0.01). PSP improved more with risperidone than with aripiprazole at week 4 and 8 (p < 0.05). Conclusions: For FES, risperidone might be a better choice than aripiprazole due to improved efficacy and functional improvement, without inferior tolerability. Aripiprazole is a better choice to avoid relevant short-term weight gain. Olanzapine could be chosen to avoid neurological adverse effects.