Background and aims Previous research has found diacetylmorphine, delivered under supervision, to be cost‐effective in the treatment of severe opioid use disorder, but diacetylmorphine is not available in many settings. The Study to Assess Long‐term Opioid Maintenance Effectiveness (SALOME) randomized controlled trial provided evidence that injectable hydromorphone is non‐inferior to diacetylmorphine. The current study aimed to compare the cost‐effectiveness of hydromorphone directly with diacetylmorphine and indirectly with methadone maintenance treatment. Design A within‐trial analysis was conducted using the patient level data from the 6‐month, double‐blind, non‐inferiority SALOME trial. A life‐time analysis extrapolated costs and outcomes using a decision analytical cohort model. The model incorporated data from a previous trial to include an indirect comparison to methadone maintenance. Setting A supervised clinic in Vancouver, British Columbia, Canada. Participants A total of 202 long‐term street opioid injectors who had at least two attempts at treatment, including one with methadone (or other substitution), were randomized to hydromorphone (n = 100) or diacetylmorphine (n = 102). Measurements We measured the utilization of drugs, visits to health professionals, hospitalizations, criminal activity, mortality and quality of life. This enabled us to estimate incremental costs, quality‐adjusted life years (QALYs) and cost‐effectiveness ratios from a societal perspective. Sensitivity analyses considered different sources of evidence, assumptions and perspectives. Findings The within‐trial analysis found hydromorphone provided similar QALYs to diacetylmorphine 0.377, 95% confidence interval (CI) = 0.361–0.393 versus 0.375, 95% CI = 0.357–0.391, but accumulated marginally greater costs $49 830 ($28 401–73 637) versus $34 320 ($21 780–55 998). The life‐time analysis suggested that both diacetylmorphine and hydromorphone provide more benefits than methadone 8.4 (7.4–9.5) and 8.3 (7.2–9.5) versus 7.4 (6.5–8.3) QALYs at lower cost $1.01 million ($0.6–1.59 million) and $1.02 million ($0.72–1.51 million) versus $1.15 million ($0.71–1.84 million). Conclusions In patients with severe opioid use disorder enrolled into the SALOME trial, injectable hydromorphone provided similar outcomes to injectable diacetylmorphine. Modelling outcomes during a patient's life‐time suggested that injectable hydromorphone might provide greater benefit than methadone alone and may be cost‐saving, with drug costs being offset by costs saved from reduced involvement in criminal activity.