Linear friction welding (LFW) is a solid-state joining process that presents significant challenges for simulation due to intense frictional heating and localised plastic deformation. Accurate modelling requires capturing thermomechanical coupling and resolving fine details of the deformation close to the weld line. Such large deformations lead to the distortion of the underlying finite element mesh, commonly leading to the use of remeshing and adaptive Lagrangian Eulerian (ALE) techniques. Conventional approaches to computational plasticity, due to the nested Newton iterations at each integration point, make the use of such features difficult due the inability to consistently map plastic history between meshes. The mapping of plastic history is particularly problematic in the finite deformation regime. As such, the focus on this work will be on evolving meshes under extreme deformations. This work extends the multifield plasticity formulation [1]. The approximation of plastic variables as fields on the finite element mesh enables thermodynamically consistent mapping of plastic history during remeshing. Recent developments to enhance robustness of the formulation will be presented, followed by incorporating a mixed formulation for the thermal problem to account for adiabatic heating. As a result of the mixed formulation, all thermoplastic state variables (plastic multiplier, plastic strain, and the temperature) are approximated in L2. This differs from standard primal finite element approaches, where temperature is approximated in H1, and plastic variables are treated at integration points. Due to the block structure of the resulting system of equations, a scalable model suitable for both CPU and emerging GPU architectures is attained, capable of utilising novel block solvers and implicit-explicit timestepping. The formulation is implemented in the open-source finite element software MoFEM [2]. Example simulations of LFW will demonstrate capabilities of the formulation.