Objectives: To investigate whether antibiotic treatment of 6 days' duration is non-inferior to treatment for 12 days in patients hospitalized for cellulitis. Methods: This multicentre, randomized, double-blind, placebo-controlled, non-inferiority trial enrolled adult patients hospitalized for severe cellulitis who were treated with intravenous flucloxacillin. At day 6 participants with symptom improvement who were afebrile were randomized between an additional 6 days of oral flucloxacillin or placebo in a 1:1 ratio, stratified for diabetes and hospital. The primary outcome was cure by day 14, without relapse by day 28. Secondary outcomes included a modified cure assessment and relapse rate by day 90. Results: Between August 2014 and June 2017, 151 of 248 included participants were randomized. The intention-to-treat population consisted of 76 and 73 participants allocated to 12 and 6 days of antibiotic therapy, respectively (mean age 62 years, 67% males, 24% diabetics); 38/76 (50.0%) and 36/73 (49.3%) were cured in the 12- and 6-day groups respectively (ARR 0.7 percentage points, 95%CI: –15.0 to 16.3). Cure rates were 56/76 (73.7%) and 49/73 (67.1%) with the modified cure assessment (ARR 6.6, 95%CI: –8.0 to 20.8). After initial cure without relapse, day 90 relapse rates were higher in the 6-day group (6% versus 24%, p < 0.05). Conclusions: Given the wide confidence intervals, we can neither confirm nor refute our hypothesis that 6 days of therapy is non-inferior to 12 days of therapy. However, a 6-day course resulted in significantly more frequent relapses by day 90. These findings require confirmation in future studies.