Optical cone beam computed tomography (CT), using a digital camera to acquire 2D projection images, provides a fast, mechanically simple method for 3D radiation dosimetry. However, original cone beam designs had poor accuracy as a result of considerable scatter/stray light reaching the camera. Previously, our group presented a redesigned convergent light source for optical cone beam CT that considerably reduced stray light contribution and improved accuracy (Dekker et al 2016 Phys. Med. Biol. 61 2910). Here, we performed an evaluation of a newly updated commercial optical cone beam CT scanner (VistaTM, ModusQA, London, Canada) based on that design. Two different light source configurations were examined: the manufacturer’s default configuration which uses a 10 cm wide, 5 cm high diffuser light source, and a smaller, 1.5 cm diameter diffuser light source that more closely aligns with our previously described design. We imaged large volume (15 cm diameter cylinders) absorbing and scattering solution phantoms as well as a 1.25 cm diameter absorber placed within 15 cm diameter gel-like scattering phantom. Optical CT reconstructions were compared against narrow-beam measurements of attenuation made by placing an aperture in the optical CT system. Our results show that considerable stray light is present when using the manufacturer’s default configuration, as cupping artifacts and large (⩾10%) discrepancies between optical CT and narrow-beam attenuation measurements occur when imaging scattering phantoms. However, when imaging is performed using the 1.5 cm diameter source, optical CT measurements agree with narrow-beam measurements within ∼3% for both absorbing and scattering objects, as well as the small absorber in a scattering medium. Using this light source will require higher optical quality vessels than are currently provided by the manufacturer.