Background: We tested whether diazepam, a GABA-ergic drug that also inhibits brain nitric monoxide formation, improves acute stroke prognosis. Methods: 880 patients, randomized within 12 h of acute stroke, received diazepam 10 mg or placebo by rectiole, as soon as possible, followed by 10-mg tablets twice daily for 3 days. Primary outcome was independence (Rankin score <3) at 3 months; secondary outcome was complete recovery (Barthel index ≧95 or Rankin score ≤1). Results: Intention-to-treat analyses on all 849 patients with full follow-up (50.4% on diazepam): odds ratio (OR) 1.14, 95% CI 0.87–1.49 for primary endpoint, and an OR of 1.26 (0.90–1.76) for complete recovery, both favoring diazepam. Adjusted analyses for all stroke patients (843): OR 1.20 (0.87–1.65), and 1.25 (0.89–1.74), respectively, and for all infarct patients (748): OR 1.31 (0.93–1.85), and 1.46 (1.02–2.09; p = 0.037), respectively. Analyses restricted to cardioembolic infarct patients (200) showed treatment benefit for the primary outcome: OR 2.26, 95% CI 1.07–4.76, p = 0.032, and complete recovery: OR 2.65, 95% CI 1.06–6.59, p = 0.037. About one third of ischemic stroke patients had ‘any adverse event’, without any difference between treatment groups. In 95 intracerebral hemorrhage patients, frequency of pneumonia and death were higher in the diazepam group than in the placebo group: 35 and 10%, 22 and 12%, respectively. Conclusions: Although point estimates favored diazepam treatment in various analyses, our data did not confirm our primary hypothesis. Diazepam treatment seems beneficial in cardioembolic infarct patients, is safe in acute ischemic stroke, but may better be avoided in intracerebral hemorrhage.