KernelBench: Can LLMs Write Efficient GPU Kernels?

Anne Ouyang, Simon Guo, Simran Arora, Alex L Zhang, William Hu, Christopher Re, Azalia Mirhoseini
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:47356-47415, 2025.

Abstract

Efficient GPU kernels are crucial for building performant machine learning architectures, but writing them is a time-consuming challenge that requires significant expertise; therefore, we explore using language models (LMs) to automate kernel generation. We introduce KernelBench, an open-source framework for evaluating LMs’ ability to write fast and correct kernels on a suite of 250 carefully selected PyTorch ML workloads. KernelBench represents a real-world engineering environment and making progress on the introduced benchmark directly translates to faster practical kernels. We introduce a new evaluation metric $\text{fast}_p$, which measures the percentage of generated kernels that are functionally correct and offer a speedup greater than an adjustable threshold $p$ over baseline. Our experiments across various state-of-the-art models and test-time methods show that frontier reasoning models perform the best out of the box but still fall short overall, matching the PyTorch baseline in less than 20% of the cases. While we show that results can improve by leveraging execution and profiling feedback during iterative refinement, KernelBench remains a challenging benchmark, with its difficulty increasing as we raise speedup threshold $p$.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-ouyang25a, title = {{K}ernel{B}ench: Can {LLM}s Write Efficient {GPU} Kernels?}, author = {Ouyang, Anne and Guo, Simon and Arora, Simran and Zhang, Alex L and Hu, William and Re, Christopher and Mirhoseini, Azalia}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {47356--47415}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/ouyang25a/ouyang25a.pdf}, url = {https://proceedings.mlr.press/v267/ouyang25a.html}, abstract = {Efficient GPU kernels are crucial for building performant machine learning architectures, but writing them is a time-consuming challenge that requires significant expertise; therefore, we explore using language models (LMs) to automate kernel generation. We introduce KernelBench, an open-source framework for evaluating LMs’ ability to write fast and correct kernels on a suite of 250 carefully selected PyTorch ML workloads. KernelBench represents a real-world engineering environment and making progress on the introduced benchmark directly translates to faster practical kernels. We introduce a new evaluation metric $\text{fast}_p$, which measures the percentage of generated kernels that are functionally correct and offer a speedup greater than an adjustable threshold $p$ over baseline. Our experiments across various state-of-the-art models and test-time methods show that frontier reasoning models perform the best out of the box but still fall short overall, matching the PyTorch baseline in less than 20% of the cases. While we show that results can improve by leveraging execution and profiling feedback during iterative refinement, KernelBench remains a challenging benchmark, with its difficulty increasing as we raise speedup threshold $p$.} }
Endnote
%0 Conference Paper %T KernelBench: Can LLMs Write Efficient GPU Kernels? %A Anne Ouyang %A Simon Guo %A Simran Arora %A Alex L Zhang %A William Hu %A Christopher Re %A Azalia Mirhoseini %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-ouyang25a %I PMLR %P 47356--47415 %U https://proceedings.mlr.press/v267/ouyang25a.html %V 267 %X Efficient GPU kernels are crucial for building performant machine learning architectures, but writing them is a time-consuming challenge that requires significant expertise; therefore, we explore using language models (LMs) to automate kernel generation. We introduce KernelBench, an open-source framework for evaluating LMs’ ability to write fast and correct kernels on a suite of 250 carefully selected PyTorch ML workloads. KernelBench represents a real-world engineering environment and making progress on the introduced benchmark directly translates to faster practical kernels. We introduce a new evaluation metric $\text{fast}_p$, which measures the percentage of generated kernels that are functionally correct and offer a speedup greater than an adjustable threshold $p$ over baseline. Our experiments across various state-of-the-art models and test-time methods show that frontier reasoning models perform the best out of the box but still fall short overall, matching the PyTorch baseline in less than 20% of the cases. While we show that results can improve by leveraging execution and profiling feedback during iterative refinement, KernelBench remains a challenging benchmark, with its difficulty increasing as we raise speedup threshold $p$.
APA
Ouyang, A., Guo, S., Arora, S., Zhang, A.L., Hu, W., Re, C. & Mirhoseini, A.. (2025). KernelBench: Can LLMs Write Efficient GPU Kernels?. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:47356-47415 Available from https://proceedings.mlr.press/v267/ouyang25a.html.

Related Material