TensorRT-LLM

NVIDIA’s open-source library that compiles Transformer blocks into highly-optimized TensorRT engines for blazing-fast LLM inference on NVIDIA GPUs.

Visit Website

Introduction

Overview

TensorRT-LLM accelerates large-language-model inference by generating TensorRT engines with custom attention kernels, paged-KV caching, quantization (FP8/FP4/INT4/INT8) and speculative decoding.

Key Capabilities

Automatic engine generation from PyTorch checkpoints
In-flight batching and look-ahead decoding for high throughput
Multi-GPU / multi-node readiness via Triton back-end
Python & C++ runtimes with OpenAI-style API

Back

Information

Websitedocs.nvidia.com
AuthorsNVIDIA
Published date2023/10/19

More Items

Ray

2017

RISELab (UC Berkeley), Anyscale Inc.

Ray is an open-source distributed compute engine that lets you scale Python and AI workloads—from data processing to model training and serving—without deep distributed-systems expertise.

ai-development ai-framework ai-train ai-serving

OpenVINO

2018

Intel

OpenVINO is an open-source toolkit from Intel that streamlines the optimization and deployment of AI inference models across a wide range of Intel® hardware.

ai-development ai-inference ai-serving

NVIDIA Dynamo

2025

NVIDIA

NVIDIA Dynamo is an open-source, high-throughput, low-latency inference framework that scales generative-AI and reasoning models across large, multi-node GPU clusters.

ai-development ai-inference ai-serving nvidia