FlashInfer-Bench: Building the Virtuous Cycle for AI-driven LLM Systems

TL;DR

FlashInfer-Bench creates a standardized, closed-loop framework that connects kernel generation, benchmarking, and deployment, enabling continuous improvement of AI-generated GPU kernels for large language model inference systems.

Contribution

It introduces a unified schema, curated dataset, benchmarking framework, and deployment mechanism to enhance AI-generated GPU kernels in LLM inference systems.

Findings

Evaluates performance and limitations of LLM agents in GPU kernel generation

Provides insights into trade-offs among GPU programming languages

Establishes a reproducible pathway for deploying AI-generated kernels

Abstract

Recent advances show that large language models (LLMs) can act as autonomous agents capable of generating GPU kernels, but integrating these AI-generated kernels into real-world inference systems remains challenging. FlashInfer-Bench addresses this gap by establishing a standardized, closed-loop framework that connects kernel generation, benchmarking, and deployment. At its core, FlashInfer Trace provides a unified schema describing kernel definitions, workloads, implementations, and evaluations, enabling consistent communication between agents and systems. Built on real serving traces, FlashInfer-Bench includes a curated dataset, a robust correctness- and performance-aware benchmarking framework, a public leaderboard to track LLM agents' GPU programming capabilities, and a dynamic substitution mechanism (apply()) that seamlessly injects the best-performing kernels into production LLM engines such as SGLang and vLLM. Using FlashInfer-Bench, we further evaluate the performance and limitations of LLM agents, compare the trade-offs among different GPU programming languages, and provide insights for future agent design. FlashInfer-Bench thus establishes a practical, reproducible pathway for continuously improving AI-generated kernels and deploying them into large-scale LLM inference.