NVIDIA releases TensorRT Model Connect in public preview

NVIDIA released TensorRT Model Connect in public preview, enabling conversion of supported Hugging Face or local checkpoints into end-to-end TensorRT inference bundles in two commands without an intermediate ONNX export step. The resulting bundles run via native C++ task APIs for generation, transcription, embedding and related workloads. The open-source project, partially built with OpenAI Codex agents under human oversight, aims to simplify high-performance model deployment across NVIDIA platforms.