Objective: To assess the cost effectiveness of a community based exercise programme as a population wide public health intervention for older adults. Design: Pragmatic, cluster randomised community intervention trial. Setting: 12 general practices in Sheffield; four randomly selected as intervention populations, and eight as control populations. Participants: All those aged 65 and over in the least active four fifths of the population responding to a baseline survey. There were 2283 eligible participants from intervention practices and 4137 from control practices. Intervention: Eligible subjects were invited to free locally held exercise classes, made available for two years. Main outcome measures: All cause and exercise related cause specific mortality and hospital service use at two years, and health status assessed at baseline, one, and two years using the SF-36. A cost utility analysis was also undertaken. Results: Twenty six per cent of the eligible intervention practice population attended one or more exercise sessions. There were no significant differences in mortality rates, survival times, or admissions. After adjusting for baseline characteristics, patients in intervention practices had a lower decline in health status, although this reached significance only for the energy dimension and two composite scores (p<0.05). The incremental average QALY gain of 0.011 per person in the intervention population resulted in an incremental cost per QALY ratio of €17 174 (95% CI = €8300 to €87 120). Conclusions: Despite a low level of adherence to the exercise programme, there were significant gains in health related quality of life. The programme was more cost effective than many existing medical interventions, and would be practical for primary care commissioning agencies to implement.