The commissioning of a Pinnacle(3) treatment planning system is described. Four Elekta linear accelerators were commissioned for external beam photons. Measured data were used to derive parameter values for the Pinnacle(3) beam model by (1). fitting a Monte Carlo model of the accelerator head to measured data and then extracting the parameters for the Pinnacle(3) beam model, and by (2). using the auto-modelling facility within Pinnacle(3). Both of these methods yielded dose distributions in accord with published recommendations. A separate small-field beam model, customized for an in-house compact blocking system, was also created, which satisfied appropriate acceptance criteria for stereotactically guided conformal brain treatments. Inhomogeneous, oblique, asymmetrical and irregular fields were also assessed, with calculated and measured doses agreeing to within +/-3%. Dose-volume histogram calculation was found to be accurate to within +/-5% dose or volume for a grid size of 4 mm x 4 mm x 4 mm, with better accuracy being achieved for finer grids. Isocentric doses were compared between Pinnacle(3)'s collapsed cone convolution algorithm and the Bentley-Milan algorithm within the Target-2 treatment planning system. Dose differences were generally less than 3% in the dose prescribed, with larger values for breast plans, where the Pinnacle(3) algorithm calculated scatter more accurately. Pelvic and thoracic plans were also verified using an anthropomorphic phantom, with local dose differences between calculated and delivered dose of up to 8%, but mainly less than 3%, and with no systematic difference. Ionization chamber verifications using START and RT-01 trial procedures demonstrated differences between calculated and measured doses of less than 2%. Following satisfactory performance in the commissioning process, Pinnacle(3) has now been introduced into routine clinical use.