Permeability is one of the key parameters affecting the development of coal-bed gas. The logging method can be used to obtain the reservoir permeability comprehensively,however,the accuracy of the logging method mainly de- pends on the computation model of the relationship between pore size and permeability. At present,computation models for gas permeability are mostly based on the assumption that the pore size is the same,ignoring the influence of the dis- tribution of pore size. For conventional reservoirs with uniform pore distribution,the error of permeability between pred- icated value and the measured value can be ignored. However,for low permeability reservoirs with complex pore sizes, the distribution of pore size has tremendous influence on permeability,especially gas permeability,the error is very large that cannot be ignored. The relationship between the distribution characteristics of pore size and gas permeability of No. 8 coal seam in the Gujiao coalmine of Shanxi Xishan Coal Mining Group was studied by MesoMR23-060H-I, NMR spectrometer manufactured by Shanghai Newmai Company,and the device of gas perme-ability developed by Taiyuan University of Technology independently. Then the computation model of gas permeability considering pore size distribution was established. The difference of gas permeability between measured values,calculated by the new model and calculated by the traditional models was compared. Taking the Gaussian distribution of pore size as an example, the effect of distribution of pore size on gas permeability was studied. The results indicated that the calculated results by the new model was in more agreement for the measured gas permeability than that of the traditional models. Assume that the coal samples obtained from low permeability coal seam with same porosity,gas pressure and distribution expec- tation,and the pore size satisfied the Gaussian distribution,when the standard deviation was larger,the pore structure was more complicated,the gas permeability was larger,and the difference of gas permeability was larger compared with computation models. When the standard deviation increased from 0. 05 to 0. 18,the gas permeability calculated with pore size distribution was changed from 3． 97×10-15 m2 to 4. 2×10-15 m2 ,with an increase of 10. 7% ,and the difference changed from 0. 97% to 11. 78% compared with computation models. Assuming that coal samples were obtained from low permeability coal seam with same porosity and standard deviation,and the pore size satisfied the Gaussian distribu- tion,when distribution expectation was larger,the gas permeability was larger,the gas permeability calculated by the new model is less different from traditional models. When the expectation changed from 0. 35 to 0. 55,the gas permea- bility calculated with pore size distribution was changed from 2. 18 ×10-15 m2 to 4. 86 ×10-15 m2 ,with an increase of 123% . When calculating the permeability of low permeability coal seams with complex pore size,the new model can calculate the gas permeability more accurately.