Fuelled in particular by current medical challenges in Traumatic Brain Injury, the field of brain mechanics has progressed tremendously in the last decade. Simulations that required months in development and runtime can now be created from patient specific medical images, meshed and run in a fraction of that time. However, and paradoxically, not only do these new models still lack multiphysics capabilities – in clear contradiction with their ultimate purpose to predict functional alteration – but their high fidelity has also exacerbated a problem that the rough early finite element models were naturally avoiding by design, i.e., the stochasticity of the brain. In this presentation, we first highlight why multiphysics modelling considerations are indeed required to adequately tackle the current challenges of brain engineering research. We then focus on the stochasticity of the brain by proposing a series of new numerical methods built on the finite element method accounting for both the intrinsic stochasticity (i.e., variations of the properties within the brain) and extrinsic stochasticity (i.e., variations between individuals). Relevant applications are given throughout as illustrations.