Herbivores play a key role in the carbon (C) cycle of arctic ecosystems, but these effects are currently poorly represented within models predicting land–atmosphere interactions under future climate change. Although some studies have examined the influence of various individual species of herbivores on tundra C sequestration, few studies have directly compared the effects of different herbivore assemblages. We measured peak growing season instantaneous ecosystem carbon dioxide (CO2) exchange (photosynthesis, respiration and net ecosystem exchange) on replicated plots in arctic tundra which, for 14 years, have excluded different portions of the herbivore population (grazed controls, large mammals excluded, both small and large mammals excluded). Herbivory suppressed photosynthetic CO2 uptake, but caused little change in ecosystem respiration. Despite evidence that small mammals consume a greater portion of plant biomass in these ecosystems, the effect of excluding only large herbivores was indistinguishable from that of excluding both large and small mammals. The herbivory-induced decline in photosynthesis was not entirely attributable to a decline in leaf area but also likely reflects shifts in plant community composition and/or species physiology. One shrub species – Betula nana – accounted for only around 13% of total aboveground vascular plant biomass but played a central role in controlling ecosystem CO2 uptake and release, and was suppressed by herbivory. We conclude that herbivores can have large effects on ecosystem C cycling due to shifts in plant aboveground biomass and community composition. An improved understanding of the mechanisms underlying the distinct ecosystem impacts of different herbivore groups will help to more accurately predict the net impacts of diverse herbivore communities on arctic C fluxes.