Powered by Kwaipilot/KAT-Dev-72B-Exp-FP8, a large language model. This application uses Hugging Face ZeroGPU for highly efficient inference.