The semi-Analytical model for predicting electrical conductivity of carbon nanotubes based composites developed by Deng and Zheng  was extended to high temperatures. To be able to apply such model to temperatures up to 800 C, correlations coefficients were examined under the thermogravimetric analysis perspective. Four different heating rates (10 Co/min, 20 Co/min, 40 Co/min, and 50 Co/min) and four CNT/epoxy compositions (control sample, 0.5 wt. %, 1.0 wt. %, and 2.0 wt. %) were investigated. As carbon nanotubes have the tendency to agglomerate a morphological study was performed using Atomic Force Microscopy. The investigation revealed no major agglomerations and a good spatial dispersion. The TGA data were treated considering the reacted fraction as function of the absolute temperature. Two different coefficients were proposed based on reacted fraction, activation energy, and the rate of reaction. The semi-Analytical expressions are physically related to the thermal degradation phenomenon, including the carbon nanotubes de-percolation effect. The model is continuous on temperature and reacted fraction, but it is discrete on CNT concentration. The model seems to capture the physical phenomenon of thermal decomposition all three main phases with accuracy.