tensorrt-cu13-libs

What it is and what it does

tensorrt-cu13-libs is NVIDIA's inference optimization and execution engine for deep learning models on CUDA-capable GPUs. It provides compiled libraries and Python bindings to convert trained neural networks (from ONNX, PyTorch, TensorFlow, or direct API definition) into optimized TensorRT engines that execute with lower latency and higher throughput than native frameworks.

Version 11.X represents a major API redesign that removes legacy features: weakly-typed networks are replaced by strongly-typed networks, implicit quantization by explicit quantization, IPluginV2 by IPluginV3, and Python support now requires Python 3.10 or newer. The package is intended for developers building production inference services, embedded AI applications, and real-time inference pipelines where GPU acceleration and model optimization are critical.

Use it for:

• Optimize and deploy large language models (LLMs) for low-latency inference on NVIDIA GPUs in production.
• Convert ONNX or PyTorch models to TensorRT engines with quantization for edge deployment and reduced memory footprint.
• Build real-time computer vision inference pipelines using optimized TensorRT engines from vision model checkpoints.
• Benchmark and profile deep learning models to measure inference performance and identify optimization opportunities.
• Integrate GPU-accelerated inference into microservices or Triton Inference Server deployments for scalable model serving.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Provides NVIDIA TensorRT libraries for GPU-accelerated deep learning inference, packaged for CUDA 13 environments.

Yes, with conditions. Install if you need GPU-accelerated inference on CUDA 13 systems and can meet the system prerequisites (CUDA toolkit, Python 3.10+, compatible GPU). High install friction and proprietary licensing are trade-offs for significant inference performance gains. Not suitable if you lack CUDA infrastructure, require Python 3.9 support, or need open-source licensing.

Install

tensorrt-cu13-libs on PyPI

pip install tensorrt-cu13-libs

uv add tensorrt-cu13-libs

poetry add tensorrt-cu13-libs

Installing tensorrt-cu13-libs

Before you install

High install friction due to large compiled binary dependencies (tensorrt_cu13_libs-11.2.1.2.tar.gz). Package is actively maintained with recent releases, but installation requires CUDA 13 and system-level prerequisites.

License in practice

Licensed under a proprietary license with unclear treatment. Users should review NVIDIA's licensing terms before deployment, particularly for commercial or production use cases.

Quickstart

pip install tensorrt-cu13-libs

import tensorrt as trt

logger = trt.Logger(trt.Logger.WARNING)
builder = trt.Builder(logger)

Requires CUDA 13.3 or 12.9 and compatible NVIDIA GPU drivers; Python 3.10 or newer (3.9 and older no longer supported as of version 11.X).

Verify before relying

• Whether this package is the prebuilt Python distribution or the OSS build components—the description mentions both but it is unclear which this PyPI entry represents.
• Specific NVIDIA GPU compute capability requirements and whether all NVIDIA GPUs are supported.
• Whether cuDNN is required at runtime or only for building from source.

Package facts

Evidence: tensorrt_cu13_libs-11.2.1.2.tar.gz

Keywords: nvidia, tensorrt, deeplearning, inference

Tags

More Artificial Intelligence packages

Further reading