Listening is a major challenge for many English-as-a-foreign language (EFL) learners. Decoding training, which helps learners develop the ability to recognize words from speech, is frequently used to assist EFL learners. Although recent empirical studies on decoding training have provided positive evidence on its effectiveness in improving EFL listening proficiency, our knowledge about the specific design characteristics of an effective decoding training model is still limited. Based on a comprehensive review, this study proposed a theory-based flipped SEF-ARCS decoding model consisting of three major components: (a) design principles for flipped learning, (b) the SEF-Automation (suitability, explore, feedback, generalization, and automation) decoding principles and (c) the ARCS (attention, relevance, confidence, satisfaction) motivational model. This study then empirically tested the effect of the flipped SEF-ARCS decoding model and found that students using the model (N = 44) performed significantly better than their counterparts learning without the model (N = 36) in terms of decoding skills and listening proficiency. Students' perceptions were also explored.