Water management is one of the most important operations in greenhouse baby leaf production. However, growers mainly irrigate the plants based on experience, which generally leads to yield loss, uneven quality, and low water-use efficiency. This study evaluated four evapotranspiration (ET) models, such as Radsum, Penman methods, FAO Penman-Monteith, and Priestley-Taylor, for irrigation strategy by predicting the ET level of greenhouse baby pakchoi [Brassica rapa L. ssp. chinensis (L.) Hanelt] under different plant densities (72-, 128-, 200-, and 288-plug tray). Among environmental factors, net radiation and photosynthetically active radiation (PAR) had the highest correlation with ET, with R2 of 0.93 and 0.94, respectively. Plant growth period was divided into different stages according to canopy development and substrate surface coverage. The corresponding crop coefficient (Kc) was introduced into ET prediction models. The result shows overestimation of ETc (crop evapotranspiration) by the Radsum and Penman methods. FAO Penman-Monteith and Priestley-Taylor methods performed the best with R2 ≈0.7 for all planting densities. These two methods are recommended for greenhouse irrigation scheduling in baby pakchoi production.