A model-predictive software sensor was developed for on-line estimation of monomer conversion and average molecular weight during bulk polymerization of systems exhibiting a gel effect. The viscosity and temperature of the reaction mass are the measured secondary variables, which when used with the model allow the state of the system to be estimated. The viscometer-reactor assembly was modified so as to measure the viscosity of the reaction mass during bulk polymerization of methyl methacrylate (MMA) at temperatures higher than those reported in our earlier work (Mankar, R. B.; Saraf, D. N.; Gupta, S. K. Ind Eng Chem Res 1998, 37, 2436). The viscosity data were curve-fitted using the modified Martin equation. Optimal temperature histories were then computed off-line, using a genetic algorithm, and implemented on the viscometer-reactor assembly in which the bulk MMA polymerization was carried out. The fact that the model tuned with the data obtained under the isothermal reactor operation can be used to predict the viscosity for nonisothermal (optimal or otherwise) reactor conditions without further tuning establishes the efficacy of the software sensor. This study can now be extended to investigate, experimentally, the on-line optimizing control of bulk MMA polymerizations.