ABSTRACT The commonly used SAB, SABC and run-of-mine (ROM) ball mill circuits can achieve high throughputs with good operating stability. However, these circuits consume a substantial amount of grinding media, which can have a substantial impact on the OPEX and the ‘carbon footprint’, specifically when viewed as an ‘embodied’ energy input. On the other hand, fully autogenous (AG) circuits can operate at much lower operating costs, but with limited throughput. The control of AG mills, fed by ores with hard and soft components, can also be challenging, due to the sensitivity of the mill to the ratio of hard to soft components. In this paper, a multi-component modelling approach was used to demonstrate that the harder component should be used as grinding media, and its use should be kept to a minimum to avoid reduction in the mill throughput. There is an optimal blend of hard to soft, and any optimal blending condition is both ore-dependent and mill-specific. Establishing this blend is challenging, and consequently discourages the use of autogenous mills. The new modelling approach addresses this challenge by providing a quantitative prediction of the influence of blending soft and hard components in a mill. This prediction can then be used to establish the required blend for maintaining a stable and efficient operation in autogenous grinding mode.