In response to the problem of high complexity and low accuracy in hybrid-field channel estimation faced by eXtremely Large-scale Multiple-Input Multiple-Output （XL-MIMO） systems in 6th Generation wireless communication technology （6G） networks， an Adaptive Gradient Matching Pursuit （AGMP） algorithm was proposed. Firstly， the angular-domain transformation matrix was used to estimate far-field components， and the polar-domain transformation matrix was used to estimate near-field components， thereby transforming the channel estimation problem into a sparse reconstruction problem. Then， during the component estimation process， the Least Mean Square （LMS） algorithm was combined with an adaptive gradient search strategy to optimize path component estimation through dynamic adjustment of step-size parameters， and the Minimum Mean Squared Error （MMSE） target was approximated iteratively， thereby optimizing the channel estimation process. Finally， the complete hybrid-field channel was reconstructed by using angular-domain and polar-domain transformation matrices， thereby achieving accurate hybrid-field channel estimation. Simulation results demonstrate that in low-Signal-to-Noise Ratio （SNR） environments， the proposed algorithm improves the achievable rate by 20.46% approximately compared to Orthogonal Matching Pursuit （OMP） algorithm. Furthermore， as the number of User Equipment （UE） antennas increases， the Normalized Mean Squared Error （NMSE） of the proposed algorithm is reduced by 1.2 dB approximately compared to that of OMP algorithm in multi-antenna environment. It can be seen that the proposed algorithm achieves superior estimation performance in low-SNR and multi-antenna UE environments.