This study aims to construct, validate, and test the effectiveness of a speed training model based on the needs of karate competitions to increase the punching and kicking speed of karate athletes aged 15–17 years. The study used a Research and Development (R&D) method that adapted the steps of Borg and Gall, including needs analysis, model development, expert validation, revision, and effectiveness testing. The needs analysis was conducted through match observations, coach interviews, training program analysis, and literature review to identify the demands of modern kumite competitions. The developed model integrates movement speed and reaction speed components designed according to the characteristics of karate competitions. Validation involved sports coaching experts, karate experts, and movement learning experts. The effectiveness testing used a one-group pretest–posttest design on 65 karate athletes in the kumite category from the INKANAS Dojo in Bone Regency and Makassar City. The results showed that the developed model met the valid criteria and was feasible to be applied in the training process. The effectiveness test results showed a significant increase in the speed of athletes' punches and kicks, indicated by a decrease in the average time from 0.8161 seconds in the pretest to 0.4398 seconds in the posttest (t = 65.240; p < 0.001). These findings indicate that the competition-based training model is able to effectively improve speed performance and provides an alternative coaching approach that is more specific and relevant to the demands of modern kumite competitions.