In spite of its capacity to capture fine-grained features, hyperspectral images (HSIs), which provide extensive spectral and spatial information, have become more important in a variety of applications. Unlike traditional images, they give a multidimensional representation that allows for improved discrimination and analysis. This work analyses the effect of optimizers on the performance of the deep learning (DL) based models for hyperspectral image classification (HSIC) in terms of overall accuracy ($𝑂𝐴$), actual accuracy ($𝐴𝐴$), and $𝐾𝑎𝑝𝑝𝑎$. The research emphasizes the critical function of optimizers in deep learning model training for hyperspectral image processing, impacting convergence dynamics and generalization proficiency. The chosen testbed for testing the influence of optimizers is HybridSN, which is well-known for its ability to combine 2D and 3D convolutions to successfully extract both spectral and spatial data. Additionally, HybridSN is the ideal choice for investigating optimizer impacts on hyperspectral image classification outcomes because to its straightforward architecture and constant performance across varied datasets such as Indian Pines (IP), Pavia University (PU), and Salinas (SA). This study juxtaposes numerous optimizers for classification tasks on different datasets, emphasizing DiffMoment as the best performer for the IP dataset, while AdamP, RAdam, and DiffMoment shine on the PU dataset, and all optimizers do well except SGD on the SA dataset.