This research is a True experimental study with a 2x2 factorial design aimed at finding the influence of the STEM approach and students' collaboration abilities on physics problem-solving skills. Sampling in this study used a simple random sampling technique with the class lottery, resulting in a sample of 31 students in the experimental class and 31 students in the control class. This study was executed by providing treatment as a STEM approach to fluid material. Data on students' collaboration abilities were collected using questionnaires, while data on physics problem-solving skills were collected using tests. The research data were analysed using Two-Way ANOVA statistical tests. The research results show that: (1) 𝐹𝑐𝑜𝑢𝑛𝑡 (625.72) > 𝐹𝑡𝑎𝑏𝑙𝑒 (3.33), meaning there are differences in physics problem-solving skills between students taught using the STEM approach and the scientific approach, (2) 𝐹𝑐𝑜𝑢𝑛𝑡 (1,793.79) > 𝐹𝑡𝑎𝑏𝑙𝑒 (3.33), meaning for high collaboration abilities, there are differences in physics problem-solving skills between students taught using the STEM approach and the scientific approach, (3) 𝐹𝑐𝑜𝑢𝑛𝑡 (1,792.37) > 𝐹𝑡𝑎𝑏𝑙𝑒 (3.33), meaning for low collaboration abilities, there are differences in physics problem-solving skills between students taught using the STEM approach and the scientific approach, (4) 𝐹𝑐𝑜𝑢𝑛𝑡 (-34,125.75) < 𝐹𝑡𝑎𝑏𝑙𝑒 (3.33), meaning there is no interaction between the STEM approach and students' collaboration abilities on physics problem-solving skills.