We report a real-world evaluation of the first commercially approved automated insulin delivery (AID) system, MiniMed 670G (670G), and open source-automated insulin delivery (OS-AID) systems. This was undertaken as a retrospective observational study in adults with type 1 diabetes using AID systems for 6 months or longer in a publicly funded health service using clinically validated data. Sixty-eight adults (38 670G, 30 OS-AID systems) were included. OS-AID system users were younger, had a shorter diabetes duration and a higher education status. OS-AID systems displayed a significantly better change in HbA1c (median À0.9% [À0.4%, À1.1%] vs. À0.1% [IQR À0.7%, 0.2%], P = .004) and time in range 3.9-10 mmol/L (mean 78.5%, SD ± 12.0% vs. 68.2% ± 14.7%, P = .024) compared with 670G. Both systems showed minimal hypoglycaemia, with OS-AID systems revealing significantly improved secondary outcomes of mean glucose and percentage of time more than 10 mmol/L, with a higher percentage of time of less than 3 mmol/L. OS-AID system users displayed improved glycaemic outcomes with no clinical safety concerns compared with 670G, although higher weight-adjusted insulin dose and weight gain were noted. The study highlights key differences in OS-AID system user characteristics that are important for interpreting real-world findings from recent OS-AID system studies.