Developed an expanded model registry for the brain-score/vision repository, focusing on integrating ConvNeXt and DeiT3 gram-aligned models to improve benchmarking coverage and deployment readiness. Leveraging expertise in computer vision, deep learning, and model deployment, the work involved updating the registry to include seven base and gram-aligned model variants, along with enriching region_layer_map metadata through over ten new JSON mappings for in1k and in22k benchmarks. The approach emphasized standardizing model naming and metadata, co-authoring changes to ensure consistency. All development was conducted in Python, prioritizing feature delivery and registry quality to enhance reproducibility and facilitate future benchmarking efforts.