Background Despite widespread use of learning collaboratives, few randomized trials have evaluated their effectiveness as a strategy for implementing evidence based practices. This randomized trial evaluated the effectiveness of a virtual learning collaborative (VLC) in the implementation of a health promotion program for persons with serious mental illness (SMI) aimed at reducing cardiovascular risk reduction in routine mental health settings, compared to routine technical assistance (TA). Methods Fifty-five mental health provider organizations were recruited to participate in a Hybrid Type 3 cluster randomized implementation-effectiveness trial of the InSHAPE health promotion program for persons with SMI. Sites were stratified by size and randomized prior to implementation to an 18-month group-based VLC with monthly learning sessions or individual site TA with four scheduled conference calls over 18 months. Primary implementation and service outcomes were InSHAPE program fidelity, participation, and reach. Primary clinical outcomes were weight loss, cardiorespiratory fitness, and cardiovascular risk reduction (>= 5% weight loss or > 50 m increase on the 6-Minute Walk Test). Program fidelity was assessed at 6, 12, and 24 months; program participation and participant-level outcomes were assessed at 3, 6, 9, and 12 months. Results VLC (N = 27) and TA (N = 28) sites were similar in organizational characteristics (all p > 0.05). At 12-month follow-up mean program fidelity score was higher in VLC compared to TA (90.5 vs. 79.1; p = 0.002), with over double the proportion with good fidelity (VLC = 73.9% vs. TA = 34.8%; p = 0.009). Over half of individuals in both VLC and TA achieved cardiovascular risk-reduction at 6-month follow-up (VLC: 51.0%; TA: 53.5%; p = 0.517) and at 12-month follow-up (62% VLC and TA; p = 0.912). At 12-month follow-up VLC compared to TA was associated with greater participation (VLC 69.5% vs. TA 56.4% attending at least 50% of sessions, p = 0.002); larger caseloads (VLC = 16 vs. TA = 11; p = 0.024); greater reach consisting of 45% greater number of participants receiving InSHAPE (VLC = 368 vs. TA = 253), and 58% greater number of participants achieving cardiovascular risk reduction (VLC = 150 vs. TA = 95). Conclusion Virtual learning collaboratives compared to routine technical assistance as an implementation strategy for evidence-based health promotion promote greater intervention fidelity, greater levels of intervention participation, greater reach, and a greater number of participants achieving clinically significant risk reduction outcomes, while achieving similarly high levels of intervention effectiveness for participants who completed at least 6 months of the program.