Robotics Education (RE) is viewed as a tool to narrow the gender gap in STEM fields, especially for promoting girls' learning. As an emerging and potential model in RE, Pair Learning (PL) is different from Individual Learning (IL), which is impacted by many factors. Among them, social factors need to be further explored. Therefore, this study emphasized gender as a key social factor and focused on girls' learning performance in different learning models (IL and PL) and gender pairing patterns (boy-girl and girl-girl), as well as exploring the effective pedagogy for enhancing girls' learning performance in RE. To this end, we conducted a comparison experiment in two classes from the fifth grade in a Chinese primary school. Results indicated that: (a) PL outperforms IL in enhancing girls' learning engagement; (b) PL and IL have the same effect on the girls' learning attitude and robotics works; (c) Mixed-gender pairing is more beneficial for girls to complete robotics works than single-gender pairing; and (d) Mixed-gender pairing and single-gender pairing have the same effects on girls' learning attitudes and engagement. The key findings, possible reasons, and implications for practice are also discussed.