Despite significant progress on the proportion of individuals who know their HIV status in 2020, Côte d'Ivoire (76%), Senegal (78%), and Mali (48%) remain far below, and key populations (KP) including female sex workers (FSW), men who have sex with men (MSM), and people who use drugs (PWUD) are the most vulnerable groups with a HIV prevalence at 5-30%. HIV self-testing (HIVST), a process where a person collects his/her own specimen, performs a test, and interprets the result, was introduced in 2019 as a new testing modality through the ATLAS project coordinated by the international partner organisation Solthis (IPO). We estimate the costs of implementing HIVST through 23 civil society organisations (CSO)-led models for KP in Côte d'Ivoire (N = 7), Senegal (N = 11), and Mali (N = 5). We modelled costs for programme transition (2021) and early scale-up (2022-2023). Between July 2019 and September 2020, a total of 51,028, 14,472, and 34,353 HIVST kits were distributed in Côte d'Ivoire, Senegal, and Mali, respectively. Across countries, 64-80% of HIVST kits were distributed to FSW, 20-31% to MSM, and 5-8% to PWUD. Average costs per HIVST kit distributed were $15 for FSW (Côte d'Ivoire: $13, Senegal: $17, Mali: $16), $23 for MSM (Côte d'Ivoire: $15, Senegal: $27, Mali: $28), and $80 for PWUD (Côte d'Ivoire: $16, Senegal: $144), driven by personnel costs (47-78% of total costs), and HIVST kits costs (2-20%). Average costs at scale-up were $11 for FSW (Côte d'Ivoire: $9, Senegal: $13, Mali: $10), $16 for MSM (Côte d'Ivoire: $9, Senegal: $23, Mali: $17), and $32 for PWUD (Côte d'Ivoire: $14, Senegal: $50). Cost reductions were mainly explained by the spreading of IPO costs over higher HIVST distribution volumes and progressive IPO withdrawal at scale-up. In all countries, CSO-led HIVST kit provision to KP showed relatively high costs during the study period related to the progressive integration of the programme to CSO activities and contextual challenges (COVID-19 pandemic, country safety concerns). In transition to scale-up and integration of the HIVST programme into CSO activities, this model shows large potential for substantial economies of scale. Further research will assess the overall cost-effectiveness of this model.