Taylor bubbles are characteristic large-scale bubbles present in the slug flow regime. They exhibit physically rich behaviour in different background flow situations and, thus, are often subject of academic research. In the laminar flow the Taylor bubble is axisymmetric, however, with sufficiently large pipe diameter and, consequently, higher Reynolds number, the shape of the bubble’s nose may become asymmetric, in particular when the bubble is exposed to counter-current liquid flow. This somewhat non-intuitive behaviour is further complicated when instabilities arise at large Reynolds number, leading to breakup at the tip of bubble trailing edge. We have investigated the Taylor bubble behaviour in a counter-current water flow, which exactly balances the buoyancy force of the bubble and allows observations at a fix location in the pipe for several hours. A Taylor bubble in counter-current flow has been reproduced with wall-resolved Large Eddy Simulation (LES) approach using OpenFOAM. We applied a modified interFoam solver, which solves the incompressible Navier-Stokes equations together with the Volume of fluid (VoF) method for interface tracking. In the present study, the model is further improved with application of Piecewise Linear Interface Calculation (PLIC) geometric reconstruction of the gas-liquid interface and a second-order implicit Runge-Kutta time-integration scheme. An adaptive flow rate at the inlet allows us to keep the Taylor bubble at fixed position in the pipe. These simulations turned out to be much more accurate than previous simulations, which applied the algebraic VoF method instead of the our geometric one. Moreover, the results revealed an interesting flow behaviour with sensitive flow structures that we were not able to detect before. In particular, a large toroidal vortex inside the Taylor bubble, which is driven by the surrounding water flow, and a secondary toroidal vortex observed inside the Taylor bubble at its trailing edge. In the liquid water region, two toroidal vortices are observed as well: a primary vortex at the bubble wake and a secondary vortex further downstream. The latter is weaker and we were able to predict it only on our finest meshes, however, it was confirmed with Particle Image Velocimetry (PIV) measurements. Our work aims to develop improved models for the gas-water interface behaviour, which are relevant for many industrial devices and applications including nuclear power plants.