As the one kind of dominant component under broad-leaves Korean pine forest stands, the early-summer herbs have an important influence on herb layer species diversity and biomass in the whole growth periods. Studying on the biomass allocation structure and biomass allometric model in different growth stages will further help to understand the survival strategy and carbon storage of these species. Cardamine leucantha, Paris verticillata, Smilacina japonica, Meehania fargesii, Anisodus acutangulus and Sanicula rubriflora were selected for phenological observation and periodical sampling from the beginning of April to the end of August. Biomass allocation characteristics and the relationships between above-ground biomass (AGB) and below-ground biomass (BGB) in different growth stages for the six early-summer herbs were analyzed. With height classes as the independent variable, the single-species and mix-species allometric models were established with five model forms (linear regression of one-variable, quadratic model, exponential model, power model, and logarithmic model). Then the optimization models were chosen and validated. The result shows that the florescence generally began in late April and ended in mid June, and the fruit period began in late May and ended in mid August for the six herbs. The duration of flower and fruit period was different in different species. The plant height, biomass and root shoot ratio (R/S) were significant various with plant growth in the whole growth periods, but the changing tendency was different. The allocation proportion of components in different species was different and the proportion of biomass allocation for reproduction was usually less than 5%. The relationships between above- and below-ground biomass were both allometric relationship (the P-value less than 0.0001 and the allometric exponent a≠1). The power model was the most frequently chosen as the optimization model base on the R² (coefficient of determination) and SEE (standard error of estimate), followed by quadratic model and exponential model. The higher R² value and lower SEE value of all optimization models indicate that the model was the better usability. And the models about AGB and TB were better than BGB, the single-species models were better than the mix-species ones. The RS, EE and RMA value of all optimization models were less than 30% excepting the model of mix-species model about BGB (RMA=30.679%) and the P-value were more than 80% by data verification. We conclude that these optimization models were able to calculate biomass of early-summer herbs under canopy of broad-leaves Korean pine in this region.