Novel "differentiated service delivery" models for HIV treatment that reduce clinic visit frequency, minimize waiting time, and deliver treatment in the community promise retention improvement for HIV treatment in Sub-Saharan Africa. Quantitative assessments of differentiated service delivery (DSD) feature most preferred by patient populations do not widely exist but could inform selection and prioritization of different DSD models. We used a discrete choice experiment (DCE) to elicit patient preferences of HIV treatment services and how they differ across DSD models. We surveyed 18+year-olds, enrolled in HIV care for ≥6 months between February-March, 2019 at four facilities in Kisumu County, Kenya. DCE offered patients a series of comparisons between three treatment models, each varying across seven attributes: ART refill location, quantity of dispensed ART at each refill, medication pick-up hours, type of adherence support, clinical visit frequency, staff attitude, and professional cadre of person providing ART refills. We used hierarchical Bayesian model to estimate attribute importance and relative desirability of care characteristics, latent class analysis (LCA) for groups of preferences and mixed logit model for willingness to trade analysis. Of 242 patients, 128 (53.8%) were females and 150 (62.8%) lived in rural areas. Patients placed greatest importance on ART refill location [19.5% (95% CI 18.4, 10.6) and adherence support [19.5% (95% CI 18.17, 20.3)], followed by staff attitude [16.1% (95% CI 15.1, 17.2)]. In the mixed logit, patients preferred nice attitude of staff (coefficient = 1.60), refill ART health center (Coeff = 1.58) and individual adherence support (Coeff = 1.54), 3 or 6 months for ART refill (Coeff = 0.95 and 0.80, respectively) and pharmacists (instead of lay health workers) providing ART refill (Coeff = 0.64). No differences were observed by gender or urbanicity. LCA revealed two distinct groups (59.5% vs. 40.5%). Participants preferred 3 to 6-month refill interval or clinic visit spacing, which DSD offers stable patients. While DSD has encouraged community ART group options, our results suggest strong preferences for ART refills from health-centers or pharmacists over lay-caregivers or community members. These preferences held across gender&urban/rural subpopulations.