In advanced wireless communication systems, multiantenna mobile handsets are developed for high data rate transmissions. In compliance with international standards and guidelines for human exposure to radio-frequency (RF) fields from these devices, it is necessary to evaluate the maximum specific absorption rate (SAR) of them. In this paper, we present low-error estimation models, which can be useful for fast and accurate determinations of the maximum SAR of a multi-antenna mobile handsets. Fundamentals of the estimation models, built from physical and mathematical perspectives, are first presented and analyzed. For validating the effectiveness of the models, we conduct several SAR measurements for a simple two-antenna mobile handset prototype. By using the estimation models, we perform SAR estimations and compare the estimated SAR with measured data. It is confirmed that the proposed low-error estimation models can provide more accurate estimated SARs, compared with the estimation techniques in previous studies. For example, the maximum estimation error caused by a conventional technique of about 3% can be decreased to 1% by the proposed estimation technique with an estimation order of three.