Greenhouse gas (GHG) emissions and energy efficiency are key parameters to consider more sustainable farming systems. Management practices on farms can either improve or deteriorate energy efficiency and GHG emissions. It is therefore essential to identify relevant agricultural practices. This study determined the energy and GHG ‘eco’- efficiency of cattle breeding, using energy balance and GHG emissions of 31 farms in two breeding area in the Brazilian Amazon, using a diagnostic tool (AgriClimateChange Tool) adapted to the region. The analysed dairy and meat farms are representative of the diversity of livestock systems in the region: suckling herd, fattener and suckling-fattener and of their degree of intensification. The energy efficiency is on average 16.27 GJ/t produced live weight/year (min= 2.35 GJ/t, max= 45.45 GJ/t) which is 47% lower than studies conducted in mainland France for similar systems. This increased efficiency could be linked to pasture feeding which makes better use of natural resources (highly productive grasses and edaphoclimatic factors) compared to other systems in the world which use more inputs and infrastructure. Results showed that fertilisers/amendments and the purchase of young animals accounted for high percentages of total energy consumption and that fuel constituted the majority of the direct energy used on farms. GHG emissions are on average 24,47 teqCO 2/t produced live weight/year (min= 3.78 teqCO 2/t, max= 65.59 teqCO 2/t) more importantthan in mainland France (14.20 teqCO2 /t produced LW/year) and similar to French Guiana (30.00 teqCO2 /t produced LW/year). These farms can be also considered as a carbon sink when the forests on their surface are included in the balance, with a production cost in terms of emissions close to neutral. This study shows the interest of this diagnostic tool, which has been calibrated for the tropics, and opens up prospects for the adoption of this tool as an indicator of ‘eco’-efficiency of livestock farming systems.