Due to climate change, human activities and natural disturbances in high-latitude permafrost and seasonally frozen areas are gradually increasing, attracting more attention from scholars. However, research primarily focuses on soil biology and chemistry in these regions, with limited exploration of their mechanical properties, especially compression properties. This study aims to evaluate the effects of gravel content and freeze–thaw (F-T) cycles on the compression properties of coarse-grained layered forest soil from northeast China’s seasonally frozen regions, with the goal of predicting the soil’s compressive changes under heavy mechanical loads. Specifically, using uniaxial confined compression tests (UCCT) on 252 disturbed soil samples (including two soil layers: AB and Bhs; six gravel contents; and seven F-T cycles), three characteristic compression coefficients—precompression stress (σpc), compression index (Cc), and swelling index (Cs)—were measured. Additionally, scanning electron microscopy (SEM) was used to analyze the mesostructure evolution of coarse-grained gravel-bearing soil. Volume changes of samples were measured after 15F-T cycles with varying gravel contents. Results indicate non-linear effects of gravel content and F-T cycles on σpc. Gravel content below 50% positively influences σpc, while content above 50% increases soil pore content, decreasing σpc. Cc and Cs exhibit an approximately negative correlation with gravel content and initially increase followed by a decrease with more F-T cycles. Moreover, the σpc and Cc of the AB layer are higher than those in the Bhs layer, likely due to differences in clay and organic carbon contents. Notably, the observed trends differ from previous studies on other soil types such as farmland and paddy fields. This study fills a gap in understanding the compression characteristics of layered gravel-bearing forest soil in seasonally frozen regions, providing valuable insights for evaluating soil compression in both seasonally frozen and permafrost regions, and understanding mechanical vehicle-soil interactions. It also lays the theoretical groundwork and provides data support for constructing compression models of layered gravel-bearing forest soil.