The aim of the study was to verify the efficiency of two frisbee ultimate teaching models at 8-year grammar schools relative to age. In the experimental group was used a game based model (Teaching Games for Understanding) and in the control group the traditional model based on teaching techniques. 6 groups of female students took part in experiment: experimental group 1 (n=10, age=11.6), experimental group 2 (n=12, age=13.8), experimental group 3 (n=14, age =15.8), control group 1 (n=11, age =11.7), control group 2 (n=10, age =13.8) and control group 3 (n=9, age =15.8). Efficiency of the teaching models was evaluated based of game performance and special knowledge results. Game performance was evaluated by the method of game performance assessment based on GPAI (Game Performance Assessment Instrument) through video record. To verify level of knowledge, we used a knowledge test, which consisted of questions related to the rules and tactics knowledge of frisbee ultimate. To perform statistical evaluation Mann-Whitney U-test was used. Game performance assessment and knowledge level indicated higher efficiency of TGfU in general, but mostly statistically insignificant. Experimental groups 1 and 2 were significantly better in the indicator that evaluates tactical aspect of game performance - decision making (p<0.05). Experimental group 3 was better in the indicator that evaluates skill execution - disc catching. The results showed that the students of the classes taught by game based model reached partially better game performance in general. Experimental groups achieved from 79.17 % to 80 % of correct answers relating to the rules and from 75 % to 87.5 % of correct answers relating to the tactical knowledge in the knowledge test. Control groups achieved from 57.69 % to 72.22 % of correct answers relating to the rules and from 51.92 % to 72.22 % of correct answers relating to the tactical knowledge in the knowledge test.