llama.cpp crashes on windows when loading models over 60gb

trying to load llama 3.3 70b q5_k_m (62.3gb) on windows 11 with llama.cpp b4729 and it just crashes with 'failed to allocate memory' even though i have 128gb ram and 48gb free same model loads fine on linux with identical hardware.... pretty sure its a windows memory allocation bug in llama.cpp but i cant find anyone else reporting it smaller models (under 60gb) work fine on windows. anyone else hit this or is it just my setup being cursed