Aim The quantification accuracy of PET images depends on several factors, such as patient status, data acquisition protocol or image reconstruction settings. The objective of this study is to evaluate the impact of acquisition time on the quantitative accuracy of PET images, reconstructed by using different settings, for phantom and patient scans in a PET/MR system. Materials and Methods Phantom measurement of a FDG-filled NEMA IQ Body Phantom is acquired in 20min list mode with a true lesion-to-background ratio LBRtrue of 4.1. Full acquisition is split to receive sets of 10min, 5min, 3min, 2min and 1min images using reconstruction settings OSEM (4it, 21subs, 5mm FWHM Gaussian post-filtering) and PSF (4it, 21subs, no filtering). Image quality is evaluated by recovery coefficients (RC), signal-to-noise ratios (SNR) and segmented volumes (VOL), for the maximum activity and for mean activities segmented at 41%, 50% and 70% adapted to background. 8 68Ga-DOTANOC and 16 18F-FET patients with gliomas are included in the study. 10min list mode scans for one bed position are performed and reconstructed with OSEM and PSF to receive sets of 5min, 3min, 2min, 1min and 30s images. Image quality is assessed via standardized uptake values (SUV), LBR, SNR and VOL for the maximum and mean activities as for the phantom. All measures are referenced to the image with longest frame time and the relative deviation of each image to the reference image is statistically evaluated. Results For maximal reduction of acquisition time, by a factor of 20, image noise increases significantly in phantom and patient data with average increases of up to 100% for OSEM reconstructions and up to 200% for PSF. RCmax increases especially for bigger spheres (12% OSEM, 56% PSF), while RC41 shows around half of these deviation values. In patients, SUVmax increases by 8% to 10% with OSEM and by 50% to 65% in PSF, while SUV41 shows less dependence on acquisition time. SNRmax decreases similarly for phantom and 68Ga-DOTANOC data (-70% in OSEM, -80% in PSF), but less in 18F-FET data (-40% in both, OSEM and PSF). Reduction of acquisition time by a factor of 6 (OSEM) or 5 (PSF) does not yet significantly influence RCmax. With OSEM reconstruction in patient studies, scan time can be lowered to 3min in 18F-FET patients or 5min in 68Ga-DOTANOC patients without significant effect on SUVmax. Conclusion The increase of image noise with shorter scan duration induces significant increases of RCmax and SUVmax and a substantial decrease of SNRmax, all together affecting quantification accuracy adversely. Measures with segmentation at 41% threshold are affected notably less by reducing acquisition duration than maximum activity measures. We propose potential reductions of acquisition time of up to 50% without compromising image quality significantly.