The ability to identify jets containing b-hadrons is important for the high-pT physics program of a general-purpose experiment at the LHC such as ATLAS. Two robust b-tagging algorithms, JetProb and SV0, taking advantage of the impact parameter of tracks or reconstructing secondary vertices have been swiftly commissioned and used for several analyses of the 2010 data: bottom and top quark production cross-section measurements, searches for supersymmetry etc. Building on this success, several more advanced b-tagging algorithms are commissioned using 330 pb−1 of the 2011 data. All these algorithms are based on a likelihood ratio formalism to separate the signal (b-jet) from the background (light or in some cases charm jet) using input distributions from simulated events. The accuracy with which the simulation reproduces the experimental data is detailed, as well as the expected improvement in performance achieved with these new tagging algorithms