While recent breakthroughs such as AlphaFold3 have revolutionized protein and RNA structure prediction, they currently fail to reliably generate correct G-quadruplex structures from sequence alone.