Glass substrates are emerging as a key alternative to silicon and conventional organic substrates for high-density and high-performance systems due to their outstanding dimensional stability, enabling sub-5-<inline-formula> <tex-math notation="LaTeX">\mu \text{m} </tex-math></inline-formula> lithographic design rules, excellent electrical performance, and unique mechanical properties, key in achieving board-level reliability at body sizes larger than <inline-formula> <tex-math notation="LaTeX">15\times15 </tex-math></inline-formula> mm 2 . This paper describes the first demonstration of the board-level reliability of such large, ultrathin glass ball grid array (BGA) packages directly mounted onto a system board, considering both their thermal cycling and drop-test performances. To investigate board-level reliability, glass BGA packages, <inline-formula> <tex-math notation="LaTeX">18.5\times18.5 </tex-math></inline-formula> mm 2 in body size and 100 <inline-formula> <tex-math notation="LaTeX">\mu \text{m} </tex-math></inline-formula> in thickness, were first designed and fabricated with a daisy-chain pattern. The glass test vehicles were fabricated at panel level, and then BGA balled by ball drop process with SAC105 solder balls, 250 <inline-formula> <tex-math notation="LaTeX">\mu \text{m} </tex-math></inline-formula> in diameter at 400-<inline-formula> <tex-math notation="LaTeX">\mu \text{m} </tex-math></inline-formula> pitch. After singulation, the glass packages were mounted onto printed circuit boards using standard surface mount technology assembly processes, and then subjected to reliability testing through thermal cycling and drop tests following JEDEC reliability standards. The effect of the coefficient of thermal expansion (CTE) of glass was evaluated by investigating low- and high-CTE glass substrates, with the CTEs of 3.8 and 9.8 ppm/°C, respectively. While all glass packages passed 1000 thermal cycles at ?40/125 °C as predicted by thermomechanical modeling using the Engelmaier-Wild model, the fatigue life of high-CTE samples exceeded 5000 thermal cycles. In addition, 28/30 drop-test samples passed the required 40 and 200 drops on corner and inner circuits, respectively, with no clear effect of the glass CTE. The predominant failure modes were systematically identified for both reliability tests.