To address the challenge of extracting highly discriminative features in thermal infrared (TIR) tracking, we propose a novel Siamese tracker based on cross-channel fine-grained feature learning and progressive fusion. First, we introduce a cross-channel fine-grained feature learning network that utilizes masks and suppression coefficients to suppress dominant target features, allowing the model to capture more nuanced and subtle information. A channel rearrangement mechanism is incorporated to improve information flow efficiency, while channel equalization helps reduce the number of parameters. Additionally, layer-by-layer combination units are introduced to facilitate effective feature extraction and fusion, minimizing both parameter redundancy and computational complexity. To further enhance the integration of fine-grained details, the network employs feature redirection and channel shuffling strategies. Next, we introduce a specialized cross-channel fine-grained loss function that guides feature groups to focus on distinct, discriminative regions of the target. This loss function includes an inter-channel loss term designed to encourage orthogonality between channels, thereby increasing feature diversity and enabling the model to capture finer details more effectively. Extensive experiments show that our proposed tracker outperforms existing methods on the LSOTB-TIR and PTB-TIR benchmarks.