Infrared thermography (IRT) has been gaining in popularity in clinical and scientific research due to the increasing availability of affordable infrared cameras. This study aims to determine the similarity of measurement performance between three models of IRT camera during assessment of skin temperature before and after physical exercise. Three models of FLIR thermographic cameras (E60bx, Flir-One Pro LT, and C2) were tested. Thermal images were taken of the foot sole, anterior leg, and anterior thigh from 12 well-trained men, before and after a 30-min run on a treadmill. Image files were blinded and processed by three evaluators to extract the mean, maximum, and standard deviation of skin temperature of the region of interest. Time for data processing and rate of perceived effort was also recorded. Data processing was slower on the E60bx (CI95% E60 vs C2 [0.2, 2.6 min], p = 0.02 and ES = 0.6); vs. Flir-One [0.0, 3.4 min], p = 0.03 and ES = 0.6) and was associated with lower effort perception (E60 3.0 ± 0.1 vs. Flir-One 5.6 ± 0.2 vs C2 7.0 ± 0.2 points; p < 0.001 and ES > 0.8). The C2 and Flir-One cameras underestimated the temperature compared with the E60. In general, measuring mean temperature provided higher camera and examiner intra-class correlations than maximum and standard deviation, especially before exercise. Moreover, post exercise mean skin temperatures provided the most consistent values across cameras and evaluators. We recommend the use of mean temperature and caution when using more than one camera model in a study.