Fluid flow in the mold region of the continuous slab caster at Panzhihua Steel is investigated with 0.6-scale water model experiments, industrial measurements, and numerical simulations. In the water model, multiphase fluid flow in the submerged entry nozzle (SEN) and the mold with gas injection is investigated. Top surface level fluctuations, pressure at the jet impingement point, and the flow pattern in the mold are measured with changing submergence depth, SEN geometry, mold width, water flow rate, and argon gas flow rate. In the industrial investigation, the top surface shape and slag thickness are measured, and steel cleanliness including inclusions and the total oxygen (TO) content are quantified and analyzed, comparing the old and new nozzle designs. Three kinds of fluid flow pattern are observed in the SEN: ‘‘bubbly flow,’’ ‘‘annular flow,’’ and an intermediate critical flow structure. The annular flow structure induces detrimental asymmetrical flow and worse level fluctuations in the mold. The SEN flow structure depends on the liquid flow rate, the gas flow rate, and the liquid height in the tundish. The gas flow rate should be decreased at low casting speed in order to maintain stable bubbly flow, which produces desirable symmetrical flow. Two main flow patterns are observed in the mold: single roll and double roll. The single-roll flow pattern is generated by large gas injection, small SEN submergence depth, and low casting speed. To maintain a stable double-roll flow pattern, which is often optimal, the argon should be kept safely below a critical level. The chosen optimal nozzle had 45-mm inner bore diameter, downward 15 deg port angle, 2.27 port-to-bore area ratio, and a recessed bottom. The pointed-bottom SEN generates smaller level fluctuations at the meniscus, larger impingement pressure, deeper impingement, and more inclusion entrapment in the strand than the recess-bottom SEN. Mass balances of inclusions in the steel slag from slag and slab measurements show that around 20 pct of the alumina inclusions are removed from the steel into the mold slag. However, entrainment of the mold slag itself is a critical problem. Inclusions in the steel slabs increase twofold during ladle changes and tenfold during the start and end of a sequence. All of the findings in the current study are important for controlling slag entrainment.