BACKGROUND: Long-acting cabotegravir and rilpivirine administered monthly or every 2 months might address the challenges associated with daily oral antiretroviral therapy. The ATLAS-2M week 48 results showed non-inferiority of long-acting cabotegravir and rilpivirine administered every 8 weeks compared with that of every 4 weeks. In this study, we report the efficacy, safety, and tolerability results from the week 96 analysis. METHODS: ATLAS-2M is a randomised, multicentre, open-label, phase 3b, non-inferiority trial conducted in 13 countries, evaluating the safety and efficacy of maintenance treatment with intramuscular injections of long-acting cabotegravir and rilpivirine, administered every 8 weeks versus every 4 weeks, to people living with HIV-1. Virologically suppressed adults with HIV-1, either already receiving intramuscular long-acting cabotegravir and rilpivirine every 4 weeks (ie, ATLAS study rollover participants) or oral standard of care, were randomly assigned (1:1), in an unblinded fashion, to receive either intramuscular long-acting cabotegravir (600 mg) and rilpivirine (900 mg) every 8 weeks (ie, the every 8-week dosing group) or intramuscular long-acting cabotegravir (400 mg) and rilpivirine (600 mg) every 4 weeks (ie, the every 4-week dosing group). Randomisation was generated using the GlaxoSmithKline-validated randomisation software RANDALL NG (version 1.3.3). The primary endpoint at week 48 was the proportion of participants with plasma HIV-1 RNA measurements of 50 copies per mL or more (ie, the US Food and Drug Administration [FDA] Snapshot algorithm), which has been published previously. Here, we present the week 96 results: the proportion of participants with plasma HIV-1 RNA measurements of less than 50 copies per mL (FDA Snapshot algorithm), with a non-inferiority margin of -10%; the proportion of participants with plasma HIV-1 RNA measurements of 50 copies per mL or more (FDA Snapshot algorithm), with a non-inferiority margin of 4%; the proportion of participants with protocol-defined confirmed virological failure (ie, two consecutive plasma HIV-1 RNA measurements ≥200 copies per mL); safety; pharmacokinetics; and tolerability. This study is registered with ClinicalTrials.gov, number NCT03299049, and is currently ongoing. FINDINGS: Between Oct 27, 2017, and May 31, 2018, a total of 1149 participants were screened; of whom, 1049 (91%) were randomly assigned and 1045 (91%) initiated treatment (522 in the every 8-week dosing group and 523 in the every 4-week dosing group). The median age was 42 years (IQR 34-50). 280 (27%) of 1045 participants were assigned female at birth and 764 (73%) were white. At week 96 (FDA Snapshot algorithm), 11 (2%) of 522 participants in the every 8-week dosing group and six (1%) of 523 in the every 4-week dosing group had an HIV-1 RNA measurement of 50 copies per mL or more, with an adjusted treatment difference of 1·0 (95% CI -0·6 to 2·5), meeting the prespecified non-inferiority threshold of 4%; 475 (91%) of 522 participants in the every 8-week dosing group and 472 (90%) of 523 in the every 4-week dosing group maintained an HIV-1 RNA measurement of less than 50 copies per mL, with an adjusted treatment difference of 0·8 (95% CI -2·8 to 4·3), which met the prespecified non-inferiority threshold of -10%. One participant in the every 8-week dosing group met the confirmed virological failure criterion since the week 48 analysis at week 88, resulting in a total of nine participants in the every 8-week dosing group and two in the every 4-week dosing group having confirmed virological failure. No new safety signals were identified, and no treatment-related deaths occurred. Injection site reactions were the most common adverse event, occurring in 412 (79%) of 522 participants in the every 8-week dosing group and 400 (76%) of 523 in the every 4-week dosing group. Most injection site reactions were grade 1 or 2 (7453 [99%] of 7557 in both groups), with a median duration of 3 days (IQR 2-5). INTERPRETATION: Long-acting cabotegravir and rilpivirine dosed every 8 weeks had non-inferior efficacy compared with that of every 4 weeks through the 96-week analysis, with both regimens maintaining high levels of virological suppression. These results show the durable safety, efficacy, and acceptability of dosing long-acting cabotegravir and rilpivirine monthly and every 2 months as maintenance therapy for people living with HIV-1. FUNDING: ViiV Healthcare and Janssen Research & Development.