This study examines predictors of smoking cessation in tuberculosis patients with high HIV co-infection rates in a South African primary care setting. Current smokers were randomly allocated to brief motivational interviewing (n = 205) or receipt of a brief message (n = 204). Multi-level logistic regression was performed to identify predictors of sustained 3- and 6-month abstinence and 7-day point prevalence abstinence (PPA) at 1 month, with the facility as a random effect. The intervention was ineffective among smokers with high nicotine-dependence at 1 month, but was effective for all smokers over longer periods. Higher baseline self-efficacy predicted the 1-month 7-day PPA, but not sustained abstinence. HIV-positive participants' odds of sustained abstinence were about three times higher than those of their HIV-negative counterparts. Results support a more intensive motivational intervention and/or coping skills' training to increase self-efficacy and abstinence rates. Tobacco cessation services can be introduced in tuberculosis services where high HIV co-infection rates occur.