Brazil has a concentrated HIV epidemic and men who have sex with men (MSM) are disproportionately affected. Yet, no data is available on the HIV care cascade for this population. This study aimed to assess the HIV care cascade among MSM newly diagnosed through innovative testing strategies in Rio de Janeiro. Data from 793 MSM and travestites/transgender women (transwomen) tested for HIV at a non-governmental LGBT organization and a mobile testing unit located at a gay friendly venue were analyzed. A 12-month-after-HIV-diagnosis-censored cohort was established using CD4, viral load and combination antiretroviral therapy (cART) longitudinal data from those diagnosed with HIV. A cross-sectional HIV care cascade was built using this data. The relative risks of achieving each cascade-stage were estimated using generalized linear models according to age, self-declared skin-color, education, history of sexually transmitted diseases (STD), drug use and prior HIV testing. From Jan-2013 to Jan-2014, 793 MSM and transwomen were tested, 131 (16.5%) were HIV-infected. As of January 2015, 95 (72.5%) were linked to HIV care, 90 (68.7%) were retained in HIV care, 80 (61.1%) were on cART, and 50 (38.2%) were virally suppressed one year after HIV diagnosis. Being non-white (Relative risk [lower bound; upper bound of 95% confidence interval] = 1.709 [1.145; 2.549]) and having a prior HIV-test (1.954 [1.278; 2.986]) were associated with an HIV-positive diagnosis. A higher linkage (2.603 [1.091; 6.211]) and retention in care (4.510 [1.880; 10.822]) were observed among those who were older than 30 years of age. Using community-based testing strategies, we were able to access a high-risk MSM population and a small sample of transwomen. Despite universal care coverage and the test-and-treat policy adopted in Brazil, the MSM cascade of care indicates that strategies to increase linkage to care and prompt cART initiation targeted to these populations are critically needed. Interventions targeting non-white and young MSM should be prioritized.