This research aims to compare the internal consistency of reliability coefficient on classical test theory. Estimation accuracy of internal consistency reliability coefficient used several methods of the coefficient reliability formulation. The methods are Split-Half Method, Cronbach Alpha formula, and Kuder Richardson formula.  Determination of the test reliability coefficients used also some formula and then their results were compared with the results of their estimation accuracy. This research is a quantitative descriptive. Data were analyzed based on responses of national chemistry examination in Jambi province on academic year 2014/2015. The data of students answer sheets were taken using proportional stratified random sampling technique. There are 200 students’ responses from 162 schools (132 public schools and 30 private schools) in Jambi province. The form of data were dichotomy data and analyzed using Split-Half Method. Their reliabilities were analyzed using Cronbach Alpha formula and Kuder Richardson formula. Reliability criteria used consist of five conditions, they are 0.5; 0.6; 0.7; 0.8 and 0.9. The results of this research indicated that (a) the coefficient of reliability in classical test theory developed by measurement experts (using Split-Half Method, Cronbach Alpha formula and Kuder Richardson formula) have varying estimates of accuracy;  (b) average reliability coefficients have the precision estimation about of 0.78 up to 0.8; (c) the reliability coefficient using Spearman Brown formula was 0.78, with Rulon formula was 0.78, Flanagan formula was 0.77, Cronbach Alpha formula was 0.838, the KR20 formula was 0.838, and KR21 formula was 0.821.