Student science proficiency development demands sustainable and coherent learning environment support. Scholars argue that project‐based learning (PBL) is an efficient approach to promoting student science learning, compared to conventional instructions. Yet, few studies have delved into the learning process to explore how a coherent PBL system consisting of curriculum, instruction, assessment, and professional learning promotes student learning. To address the gap, this study investigated whether students' science proficiency on the three post‐unit assessments predicted their achievement on a third‐party‐designed end‐of‐year summative science test in a coherent high school chemistry PBL system aligned with the recent US science standards. The study employed a cluster randomized experimental design to test an intervention using our PBL system and only used data from the treatment group. The sample consisted of 1344 treatment students who participated in our PBL intervention and underwent the pretest and end‐of‐year summative test. Students' responses to the three post‐unit assessments were selected and rated to indicate their science proficiency. Two‐level hierarchical linear models were employed to explore the effects of students' performances of three post‐unit assessments on their end‐of‐year summative achievement, considering and controlling for student prior knowledge (i.e., pretest and prior post‐unit assessments). This study suggests two main findings. First, students' science proficiency in the three units could cumulatively and individually predict their summative science achievement. Second, students' performances on the two types of tasks (i.e., developing and using models) in the three post‐unit assessments could also be used to predict their summative science achievement. This research contributes to the field by showing that a coherent standards‐aligned PBL system can significantly and sustainably impact student science proficiency development.