The rotating accelerometer gravity gradiometer (RAGG) is regarded as a significant tool in geophysics applications such as resource exploration and gravity-assisted navigation research. Here we develop a prototype of the horizontal tensor RAGG and build a dedicated calibration platform for evaluating its performance. The error of calibration is theoretically computed and experimentally verified. The errors considered include higher-order terms of the gravity gradient error, positioning error, demodulation phase error, as well as the dimensions and density error of the gravitational mass. It has witnessed a high-accuracy gravity gradient calibration platform based on a motorized translation stage, with error values of 0.49 E (1 E =10-9 /s2) in channel Γsin and 0.24 E in channel Γcos. At an averaging time of 32 seconds, the RAGG demonstrates optimal stability, featuring a minimum Allan deviation value of 6.6 E in channel Γsin and 5.5 E in channel Γcos. And the RAGG has achieved a measurement noise floor of about 40 E/√Hz @ 0.001-0.1 Hz in both channels. For simulating engineering applications, the experimental results demonstrate a standard deviation error of only 3.2 E by employing a gravity gradient excitation step signal of 10 E. It shows that this RAGG has the potential for a wide range of applications.