Knee joint forces are determined either through mathematical modeling or by in vivo measurement using an instrumented knee prosthesis. In the model studies, significant differences exist among the results and the data for high knee flexion are few. The in vivo measurement data are available for small to moderate flexion but not for high flexion yet. We created a 2D mathematical model of the lower limb incorporating several new features such as a patello-femoral mechanism, a thigh-calf contact at high knee flexion and co-contracting muscles' force ratio, then used it to determine knee joint forces arising from high knee flexions in four kneeling conditions: rising with legs in parallel, with one foot forward, with or without arm use. With arms used, the maximum values of knee joint force decreased to about 60% of those with arms not used. When rising with one foot forward, if arms are not used, the forward leg sustains a force as large as that sustained when rising with legs parallel. By comparing our modeling methodology and results with those in the literature, we determined some of the causes of the differences in the results, thereby providing creditable data especially during high flexion of the knee.