How Fast Can I Run My VLA? Demystifying VLA Inference Performance with VLA-Perf

TL;DR

This paper introduces VLA-Perf, an analytical model for VLA inference performance, systematically studying how model design and deployment choices impact real-time inference capabilities in embodied AI tasks.

Contribution

The paper presents VLA-Perf, the first analytical performance model for VLA inference, and provides comprehensive insights into optimizing model design and deployment for real-time applications.

Findings

Inference performance is significantly affected by model scaling and architecture.

Long-context inputs and asynchronous inference influence latency.

Optimal deployment depends on hardware and network conditions.

Abstract

Vision-Language-Action (VLA) models have recently demonstrated impressive capabilities across various embodied AI tasks. While deploying VLA models on real-world robots imposes strict real-time inference constraints, the inference performance landscape of VLA remains poorly understood due to the large combinatorial space of model architectures and inference systems. In this paper, we ask a fundamental research question: How should we design future VLA models and systems to support real-time inference? To address this question, we first introduce VLA-Perf, an analytical performance model that can analyze inference performance for arbitrary combinations of VLA models and inference systems. Using VLA-Perf, we conduct the first systematic study of the VLA inference performance landscape. From a model-design perspective, we examine how inference performance is affected by model scaling, model architectural choices, long-context video inputs, asynchronous inference, and dual-system model pipelines. From the deployment perspective, we analyze where VLA inference should be executed -- on-device, on edge servers, or in the cloud -- and how hardware capability and network performance jointly determine end-to-end latency. By distilling 15 key takeaways from our comprehensive evaluation, we hope this work can provide practical guidance for the design of future VLA models and inference systems.