Background Neuraminidase inhibitors are effective for the treatment of acute uncomplicated influenza. However, there is an unmet need for intravenous treatment for patients admitted to hospital with severe influenza. We studied whether intravenous zanamivir was a suitable treatment in this setting. Methods In this international, randomised, double-blind, double-dummy, phase 3 trial, we recruited patients aged 16 years or older with severe influenza admitted to 97 hospitals from 26 countries. We randomly assigned patients (1:1:1 stratified by symptom onset ≤4 days or 5–6 days) to receive 300 mg or 600 mg intravenous zanamivir, or standard-of-care (75 mg oral oseltamivir) twice a day for 5–10 days; patients were followed up for 28 days. The randomisation schedule, including stratification, was generated using GlaxoSmithKline's RandAll software. Patients, site study staff, and sponsor were masked to study treatment. The primary endpoint was time to clinical response—a composite of vital sign stabilisation and hospital discharge—in the influenza-positive population. The trial was powered to show an improvement of 1·5 days or greater with 600 mg intravenous zanamivir. Pharmacokinetic, safety, and virology endpoints were also assessed. This trial is registered with ClinicalTrials.gov, number NCT01231620. Findings Between Jan 15, 2011, and Feb 12, 2015, 626 patients were randomly assigned to receive 300 mg intravenous zanamivir (n=201), 600 mg intravenous zanamivir (n=209), or 75 mg oral oseltamivir (n=205) twice a day; 11 patients discontinued the study before receiving any study treatment. 488 (78%) of 626 patients had laboratory-confirmed influenza. Compared with a median time to clinical response of 5·14 days in the 600 mg intravenous zanamivir group, the median time to clinical response was 5·87 days (difference of −0·73 days, 95% CI −1·79 to 0·75; p=0·25) in the 300 mg intravenous zanamivir group and 5·63 days (difference of −0·48 days, 95% CI −2·11 to 0·97; p=0·39) in the oseltamivir group. Four patients with influenza A/H1N1pdm09 in the oseltamivir group developed H275Y resistance mutations. Adverse events were reported in 373 (61%) of treated patients and were similar across treatment groups; the most common adverse events (300 mg intravenous zanamivir, 600 mg intravenous zanamivir, oseltamivir) were diarrhoea (10 [5%], 15 [7%], 14 [7%]), respiratory failure (11 [5%], 14 [7%], 11 [5%]), and constipation (7 [3%], 13 [6%], 10 [5%]). 41 (7%) treated patients died during the study (15 [7%], 15 [7%], 11 [5%]); the most common causes of death were respiratory failure and septic shock. Interpretation Time to clinical response to intravenous zanamivir dosed at 600 mg was not superior to oseltamivir or 300 mg intravenous zanamivir. All treatments had a similar safety profile in hospitalised patients with severe influenza. Funding GlaxoSmithKline.