Iris can be classified into the most reliable biometric characteristic because the variability of iris is very high. the important issue in iris biometrics recognition system is how to find the feature of iris. gabor filter is used in this paper to producuce the iris feature. this feature is called iris code. hamming distance algorithm is used to find the similarity degre between tested and referenced image. this system is tested using 300 iris images from 100 people which 2 samples as reference sample and 1 sample as expriment for the right eye, and 400 iris images from 100 users, 3 samples as reference and 1 sample as tested sample for the left eye. iris codes that is produced by gabor filter gives good performance to the iris recognition system with accuracy reached 94,758% (T = 0463, FNMR = 4500%, 0742% = FNMR) for the left iris and 92,397% (T=0470, FNMR = 6000%, FMR = 1603%) for the left iris. the expriment result also show that the performance of this system relatifely stable although the database size increased.