Due to the transient operation of short duration facilities (0.2–1.0s running time), fidelity in temperature reproduction requires both minimum steady-state errors and a frequency response above 20Hz. Even with the smallest thermocouple wire diameter (∼12μm), badly designed probes may suffer from unsteady heat conduction between wires and supports. The resulting error is often much larger than steady errors such as the effect of recovery factor. In this paper, the origins of steady and unsteady measurement errors are described and evaluated. An analytical modeling of the transient convection/conduction problem is presented. A number of probe designs are described and evaluated at different Reynolds number. The dynamic response is tested in a hot jet apparatus that generates temperature steps, with jet velocities up to 150m/s. The influence of the length/diameter ratio, the type of support and the presence of a shield on the dynamic response are addressed. The unsteady behavior of a thermocouple probe can be replicated with a combination of first order systems, which defines the transfer function of the probe. The relevant parameters are found by using an optimization routine that fits the numerical system response to the experimental response. This numerical model can be reversed to perform a frequency compensation of the measured data.