Some vendors have begun experimenting with on-device AI solutions, deploying small models locally through techniques such as quantization and pruning, though functionality is generally limited.