Problem based learning (PBL) models are commonly used to assess students problem solving abilities. However, there are several resistance to PBL syntax that can improve students' science process skills. The purpose of this study was to determine the PBL model of students' science process skills. This study is a quasi experimental research using samplings techniqe, where VA class as experimental class I, VB II as experimental class II, and VC as control class. Data analysis techniques using covariate analysis (ANACOVA) using by SPSS 23.0 for windows applications and data collection using multiple choice test. Based on hypothesis test with P <0.05 then H o is rejected and H a accepted where this means there is significant influence on the application of model PBL with average value is 87.60. Its means that PBL (problem from students) more positively than PBL (problem from teacher 82.95), and direct instruction model (average value 78.60) to the students science process skill.