There is a need for minimally invasive biomarkers that can accurately and quickly quantify radiation exposure. Radiation-responsive proteins have applications in clinical medicine and for mass population screenings after a nuclear or radiological incident where the level of radiation exposure and exposure pattern complicate medical triage for first responders. In this study, we evaluated the efficacy of the acute phase protein serum amyloid A (SAA) as a biomarker for radiation exposure using plasma from irradiated mice. Ten-week-old female C57BL6 mice received a 1–8 Gy single whole-body or partial-body dose from a Pantak X-ray source at a dose rate of 2.28 Gy/min. Plasma was collected by mandibular or cardiac puncture at 6, 24, 48 and 72 h or 1–3 weeks postirradiation. SAA levels were determined using a commercially available ELISA assay. Data was pooled to generate SAA μg/ml threshold values correlating plasma SAA levels with radiation dose. SAA levels were statistically significant over control at all exposures between 2 and 8 Gy at 24 h postirradiation but not at 6, 48 and 72 h or 1–3 weeks postirradiation. SAA levels at 1 Gy were not significantly elevated over control at all time points. Total-body-irradiated (TBI) SAA levels at 24 h were used to generate a dose prediction model that successfully differentiated TBI mice into dose received cohorts of control/1 Gy and ≥2 Gy groups with a high degree of accuracy in a blind study. Dose prediction of partial-body exposures based on the TBI model correlated increasing predictive accuracy with percentage of body exposure to radiation. Our findings indicate that plasma SAA levels might be a useful biomarker for radiation exposure in a variety of total- and partial-body irradiation settings.