In view of the defects in model of thermocouple characteristic using BP neural network (BPNN), such as lower precision, varying output, instability( after repeated training, the output may be queer), a model of thermocouple characteristic based on Generalized Regression Neural Network (GRNN) is established. The paper gives the process of model building for Ni-CrConstantan thermocouple characteristic within -270 ~ 1000℃. By means of network training, simulation and error analysis, the scope of spread parameter of neural network model for Ni-Cr Constantan thermocouple was found. When the spread is at 0.01~1.5, bigger errors appear mainly when thermo-EMF is less than 0V and greater than 75V. When it is at 0~0.01, the model has high precision and absolute error between simulation temperature and setting temperature is close to 0℃ (the mean squared error is 0.00000703~0 ℃ ). The results indicate that the model presented has a quick convergent speed in learning process, a higher accuracy and stability within a certain parameter scope. If the model is stored in CPU of an intelligent instrument, the instrument will have high accuracy without increasing hardware cost.