Purpose: To verify 3D dose distributions predicted by Eclipse treatment planning software (with AAA version 8.2.23) with those measured by polymer gel dosimeters for RapidArc intensity modulated arc radiation therapy. Methods: Treatment plans were created for RapidArc therapy of two prostate cancer patients. The daily fraction size was 180 cGy for both plans (R and E). The treatments used 10MV photon beams from a Varian iX linear accelerator. Two cylindrical phantoms (15 cm diameter and 15 cm long) containing BANG3‐PRO polymer gel were used. For dosimetric verification, QA plans were created using the CT images of the polymer gel phantom, which were obtained using 3‐mm thick slices. The phantoms were irradiated using the dose delivery techniques identical to those used for the patient treatments. After irradiation, the phantoms were scanned using a 32 spin‐echo technique with a 3T MRI scanner. The major imaging parameters were as follows: 1 mm × 1mm pixels, 45 2‐mm thick slices, and the echo spacing of 13.6 ms. The MRI images were processed to obtain 3D dose distributions using in‐house MATLAB programs. 3D dose distributions of measurements and calculations were compared using gamma index (3 % and 3 mm criteria), dose volume and dose difference diagrams, line dose profiles, and planer dose distributions. Results: Mean percentage differences between the measured and calculated doses in the 3D volumes containing PTVs were −3.1±11.9 % and 1.6±9.8 % for Plans R and E, respectively. The passing rates of the gamma index analyses were 90% and 84% for doses greater than 80% of the maximum doses of those plans. Conclusions: The current study has demonstrated a potential usefulness of polymer gel dosimetry for 3D dose verification of RapidArc plans. However, the agreement of measured and calculated doses was not satisfactory. The causes of the disagreements need to be investigated. Japanese Society of Radiological Technology.