Abstract. Large herbivore grazing has been shown to substantially alter tundra soil and vegetation properties as well as carbon fluxes, yet observational evidence to quantify the impact of herbivore introduction into Arctic permafrost ecosystems remains sparse. In this study we investigated growing season CO2 and CH4 fluxes with flux chambers on a former wet tussock tundra inside Pleistocene Park, a landscape experiment in Northeast Siberia with a 22 year history of grazing. Reference data for an undisturbed system were collected on a nearby ungrazed tussock tundra. Linked to a reduction in soil moisture, topsoil temperatures at the grazed site reacted one order of magnitude faster to changes in air temperatures compared to the ungrazed site and were significantly higher, while the difference strongly decreased with depth. Overall, both GPP (gross primary productivity, i.e. CO2 uptake by photosynthesis) and Reco (ecosystem respiration, i.e. CO2 release from the ecosystem) were significantly higher at the grazed site with notable variations across plots at each site. The increases in CO2 component fluxes largely compensated each other, leaving NEE (net ecosystem exchange) similar across grazed and ungrazed sites for the observation period. Soil moisture and CH4 fluxes at the grazed site decreased over the observation period, while in contrast the constantly water-logged soils at the ungrazed site kept CH4 fluxes at significantly higher levels. Our results indicate that grazing of large herbivores promotes topsoil warming and drying, effectively accelerating CO2 turnover while decreasing methane emissions. Our experiment did not include autumn and winter fluxes, and thus no inferences can be made for the annual NEE and CH4 budgets at tundra ecosystems.