This study presents an empirical model based on a GIS approach, which was constructed to estimate the large-scale carbon fluxes over the entire Russian tundra zone. The model has four main blocks: (i) the computer map of tundra landscapes; (ii) data base of long-term weather records; (iii) the submodel of phytomass seasonal dynamics; and (iv) the submodel of carbon fluxes. The model uses exclusively original in situ diurnal CO2 flux chamber measurements (423 sample plots) conducted during six field seasons (1993–98). The research sites represent the main tundra biome landscapes (arctic, typical, south shrub and mountain tundras) in the latitudinal diapason of 65–74°N and longitudinal profile of 63°E−172°W. The greatest possible diversity of major ecosystem types within the different landscapes was investigated. The majority of the phytomass data used was obtained from the same sample plots. The submodel of carbon fluxes has two dependent [GPP, Gross Respiration (GR)] and several input variables (air temperature, PAR, aboveground phytomass components). The model demonstrates a good correspondence with other independent regional and biome estimates and carbon flux seasonal patterns. The annual GPP of Russian tundra zone for the area of 235 × 106 ha was estimated as −485.8 ± 34.6 × 106 tC, GR as +474.2 ± 35.0 × 106 tC, and NF as −11.6 ± 40.8 × 106 tC, which possibly corresponds to an equilibrium state of carbon balance during the climatic period studied (the first half of the 20th century). The results advocate that simple regression-based models are useful for extrapolating carbon fluxes from small to large spatial scales.