Cortex.Tensorrt-LLM is a C++ inference library that can be loaded by any server at runtime. It submodules NVIDIA’s TensorRT-LLM for GPU accelerated inference on NVIDIA's GPUs.