Purpose: Low dose hyperradiosensitivity (HRS) has been observed in HGL21- and T98G human glioblastoma cells in vitro. The present study investigates whether these effects translate into improved outcome of ultrafractionated irradiation (UF) in vivo. Material and methods: T98G or HGL21 were transplanted on the hind leg of nude mice. Tumours were irradiated with UF (3 fractions of 0.4 Gy per day, interval 4h, 7 days per week) or with conventional fractionation (CF; 1 fraction of 1.68 Gy per day, 5 days per week) over 2 or 4 weeks in HGL21 and 2,4 or 6 weeks in T98G. In HGL21, graded top-up doses under clamped hypoxia were applied after 4 weeks of fractionated irradiation. Additional groups of animals were irradiated with single doses under clamp hypoxic conditions with or without whole body irradiation (WBI) before tumour transplantation. Experimental endpoints were growth delay (time to 5-fold starting volume, GD_V5) and local tumour control. Results: In T98G tumours median relative GD_V5 was 1.2 [95%C.I. 0.96; ∞] in the CF and 0.8 [0.7; 1.02] in the UF arm (p = 0.009) indicating that ultrafractionation is less efficient than conventional fractionation. The TCD₅₀ value of 33.5 Gy [22; 45] after UF was higher than TCD₅₀ of 23.6 Gy [16; 31] after CF (p = 0.15). In HGL21 the median relative GD_V5 was not significantly different between CF and UF. The top-up TCD₅₀ value of 16.1 Gy [95% C.I. 9; 23 Gy] after CF was significantly lower than the corresponding value of 33.2 Gy [23; 44] after UF irradiation (p = 0.007), indicating a higher efficacy of CF compared to UF. Conclusion: The results on human T98G and HGL21 glioblastoma do not support the hypothesis that HRS in vitro translates into improved outcome of ultrafractionated irradiation in vivo.