The aim of the present network meta-analysis is to assess the efficacy and safety across different long and short-acting analogs for the treatment of type 2 diabetes. A PubMed, EMBASE, and Cochrane Central Register of Controlled Trials databases search (20th May, 2020) for all trials with a duration ≥24 weeks comparing an analogue with another or human insulin was performed. Indirect comparisons were performed by NMA choosing glargine U100 and human regular insulin, as the reference for long- and short-acting analogues, respectively. Primary endpoints were HbA1c at 24, 52, and 104 weeks. The weighted difference in means (WDM) and Mantel-Haenzel Odds Ratio [MH-OR] with 95% Confidence Intervals (CI) were calculated for categorical and continuous variables, respectively. Fifty trials (n = 43) and 7 for basal and prandial analogues, respectively, enrolling 25,554 and 3184 patients with type 2 and 1 diabetes, respectively, were included. At NMA, detemir was less effective than glargine U-100 at 52 weeks. A significant reduction of 24-week HbA1c (WMD [IC]: -0.10 [-0.17, -0.03]%); and risk of total (MH-OR [IC]: 0.80 [0.70, 0.91]), and nocturnal hypoglycemia (MH-OR [IC]: 0.57 [0.45, 0.73]) was observed for basal analogues versus NPH insulin. At NMA, glargine U300 and degludec were associated with a significant reduction in the risk of nocturnal hypoglycemia. No significant differences across different short-acting insulin were observed. This paper supports the use of long-acting analogues, rather than NPH insulin, as basal insulin for the treatment of type 2 diabetes, without any preferences for any individual long-acting analogue over the others. The evidence on short acting analogues is limited.