Purpose We compared dexamethasone 12 versus 6 mg daily for up to 10 days in patients with coronavirus disease 2019 (COVID-19) and severe hypoxaemia in the international, randomised, blinded COVID STEROID 2 trial. In the primary, conventional analyses, the predefined statistical significance thresholds were not reached. We conducted a pre-planned Bayesian analysis to facilitate probabilistic interpretation. Methods We analysed outcome data within 90 days in the intention-to-treat population (data available in 967 to 982 patients) using Bayesian models with various sensitivity analyses. Results are presented as median posterior probabilities with 95% credible intervals (CrIs) and probabilities of different effect sizes with 12 mg dexamethasone. Results The adjusted mean difference on days alive without life support at day 28 (primary outcome) was 1.3 days (95% CrI -0.3 to 2.9; 94.2% probability of benefit). Adjusted relative risks and probabilities of benefit on serious adverse reactions was 0.85 (0.63 to 1.16; 84.1%) and on mortality 0.87 (0.73 to 1.03; 94.8%) at day 28 and 0.88 (0.75 to 1.02; 95.1%) at day 90. Probabilities of benefit on days alive without life support and days alive out of hospital at day 90 were 85 and 95.7%, respectively. Results were largely consistent across sensitivity analyses, with relatively low probabilities of clinically important harm with 12 mg on all outcomes in all analyses. Conclusion We found high probabilities of benefit and low probabilities of clinically important harm with dexamethasone 12 mg versus 6 mg daily in patients with COVID-19 and severe hypoxaemia on all outcomes up to 90 days.