<p>With the continuous improvement of global satellite navigation system (GNSS), the increase in the number of available satellites and the limited computational capability of mass-market GNSS chips or smartphones require higher time-efficiency in parameter estimation. In this contribution, we leverage the inherent sparse coupling characteristics between parameters and observations in GNSS models to propose a variable-dimension (VD) update strategy. This approach enables high computational efficiency in GNSS data processing by effectively reducing redundant numerical operations. Furthermore, the proposed VD update strategy can be integrated into various sequential Kalman filter (KF) variants, such as the standard sequential KF (SKF), the square-root covariance filter (SRCF), and the U-D factorization filter (UDF). In this work, the VD update strategy is specifically integrated into the UDF to create a variable-dimension UDF (VD-UDF) implementation. To evaluate its performance, the VD-UDF is compared against the standard UDF, as well as the KF in information form (KF-I) and the square-root information filter (SRIF), both of which are widely adopted in GNSS applications. Numerical results and time-efficiency analyses demonstrate that the VD-UDF achieves superior computational efficiency relative to standard UDF, KF-I, and SRIF nearly without compromising numerical accuracy or stability. In summary, the optimized VD-UDF offers a highly time-efficient and robust solution for real-time high-precision GNSS positioning on platforms with limited computational resources.</p>