Identifying cost-effective interventions to improve early literacy is vital to developing countries, given the importance of early literacy for an individual's future education outcomes and subsequent human capital formation. This paper presents the impact on early grade reading outcomes of two low-cost randomized interventions in Tonga: a reading instruction intervention and a community play-based activity intervention. The first intervention aims to improve early grade reading outcomes specifically; estimated impacts are approximately 0.3 standard deviation, although in some reading domains impacts are substantial, ranging from 0.6 to 0.7 standard deviation. The second intervention aims to improve school readiness and subsequently early grade reading outcomes, by providing communities with support to establish a community play-based activity. Using an instrumental variables approach, the play-based activity demonstrates positive impacts of around 0.2 standard deviation in many but not all reading domains. For the domains where a statistically significant impact is measured, the community play-based activity intervention is as at least as cost effective as the reading instruction intervention. The play-based activity intervention is shown to improve test scores by 0.21 to 0.47 standard deviation per US$100, depending on the reading test domain. The reading instruction intervention improves test scores by 0.08 to 0.34 standard deviation per US$100. These findings contribute further evidence on the effectiveness of reading instruction interventions, and possibly the first estimates of the impact of play group-type interventions on primary school reading outcomes.