USPTO Patent for AI Load Testing and Performance Benchmarking

ChangeBridge: Patent Grants - AI & Computing (G06N)

Published March 24th, 2026

Detected March 24th, 2026

Summary

The USPTO has granted a patent (US12585507B2) to Microsoft Technology Licensing, LLC for a system enabling repeatable load testing and performance benchmarking of AI models on cloud platforms. The patent covers techniques for generating load profiles and representative workloads to evaluate AI model performance and ensure quality.

View original document View source feed page

What changed

The United States Patent and Trademark Office (USPTO) has granted patent US12585507B2 to Microsoft Technology Licensing, LLC. This patent covers novel techniques for performing repeatable and iterative load testing and performance benchmarking of artificial intelligence (AI) models deployed on cloud computing platforms. The disclosed methods involve utilizing load profiles and representative workloads to evaluate AI models under various contexts, extracting performance metrics, and dynamically adjusting inputs to test diverse scenarios. The patent aims to enable the construction of quality gates for AI models to ensure consistent user experience.

This patent grant is primarily an intellectual property matter and does not impose direct regulatory obligations on entities. However, companies developing or deploying AI models, particularly those utilizing cloud infrastructure, may find the patented techniques relevant to their internal testing and quality assurance processes. Compliance officers should be aware of this patent as it pertains to AI development and performance validation, especially if their organization utilizes or licenses similar technologies, to ensure awareness of potential intellectual property considerations.

Source document (simplified)

← USPTO Patent Grants

Load testing and performance benchmarking for large language models using a cloud computing platform

Grant US12585507B2 Kind: B2 Mar 24, 2026

Assignee

MICROSOFT TECHNOLOGY LICENSING, LLC

Inventors

Sanjay Ramanujan, Rakesh Kelkar, Hari Krishnan Srinivasan, Karthik Raman, Hema Vishnu Pola, Sagar Taneja, Mradul Karmodiya

Abstract

The techniques disclosed herein enable systems to perform repeatable and iterative load testing and performance benchmarking for artificial intelligence models deployed in a cloud computing environment. This is achieved by utilizing load profiles and representative workloads generated based on the load profiles to evaluate an artificial intelligence model under various workload contexts. The representative workload is then executed by the artificial intelligence model utilizing available computing infrastructure. Performance metrics are extracted from the execution and analyzed to provide insight into various performance dynamics such as the relationship between latency and data throughput. In addition, load profiles and input datasets are dynamically adjusted to evaluate different scenarios and use cases enabling the system to automatically test the artificial intelligence model across diverse applications. Furthermore, by comparing various iterations of the artificial intelligence model, a quality gate can be constructed to enforce a consistent and high-quality user experience.