NAVSIM is a data-driven benchmarking framework for non-reactive autonomous vehicle simulation. It addresses the limitations of traditional benchmarks by combining large datasets with a non-reactive simulator to enable large-scale real-world benchmarking. Traditional metrics like average displacement error (ADE) fail to capture the interactive and multi-modal nature of driving. NAVSIM uses simulation-based metrics such as progress and time-to-collision to provide more meaningful evaluations of trajectory outputs from sensor-based driving policies. The framework includes a standardized training and evaluation split using the OpenScene dataset, which is a redistribution of nuPlan, the largest annotated public driving dataset. NAVSIM enables a new competition at CVPR 2024, where 143 teams submitted 463 entries, resulting in several new insights. On a large set of challenging scenarios, simple methods with moderate compute requirements such as TransFuser can match recent large-scale end-to-end driving architectures like UniAD. The framework is modular and can be extended with new datasets, data curation strategies, and metrics. NAVSIM provides a comprehensive tool for AV data curation, simulation, and benchmarking. It includes a set of simulation-based metrics that are well-suited for evaluating sensor-based motion planning. The framework also includes a non-reactive simulation where the evaluated policy and environment do not influence each other. This allows for open-loop metric computation while being better aligned with closed-loop evaluations than traditional displacement errors. NAVSIM's evaluation protocol is better aligned with closed-loop driving, and it benchmarks an established set of end-to-end planning baselines. The results show that simple models can match recent large-scale end-to-end driving architectures in challenging scenarios. The framework also includes a challenge where 143 teams from 13 countries developed diverse methods that competed on the proposed benchmark. The top methods ranged from multi-billion parameter vision language models to more efficient and recently overlooked approaches based on trajectory sampling and scoring. NAVSIM provides a standardized evaluation server on the open-source HuggingFace platform, which is free, has low maintenance overhead, and enables future scaling to more challenging datasets and metrics. The framework is designed to address the shortcomings of existing driving benchmarks and provide standardized but configurable simulation-based metrics for benchmarking driving policies. It improves accessibility for conducting AV research by providing downloadable challenging scenario splits and simple data curation methods. The results show that NAVSIM can serve as an accessible toolkit for AV researchers that bridges the gap between simulated and real-world driving. The framework is designed to support more datasets in the future and advocate for more open dataset releases by the community for accelerating progress in autonomous driving.NAVSIM is a data-driven benchmarking framework for non-reactive autonomous vehicle simulation. It addresses the limitations of traditional benchmarks by combining large datasets with a non-reactive simulator to enable large-scale real-world benchmarking. Traditional metrics like average displacement error (ADE) fail to capture the interactive and multi-modal nature of driving. NAVSIM uses simulation-based metrics such as progress and time-to-collision to provide more meaningful evaluations of trajectory outputs from sensor-based driving policies. The framework includes a standardized training and evaluation split using the OpenScene dataset, which is a redistribution of nuPlan, the largest annotated public driving dataset. NAVSIM enables a new competition at CVPR 2024, where 143 teams submitted 463 entries, resulting in several new insights. On a large set of challenging scenarios, simple methods with moderate compute requirements such as TransFuser can match recent large-scale end-to-end driving architectures like UniAD. The framework is modular and can be extended with new datasets, data curation strategies, and metrics. NAVSIM provides a comprehensive tool for AV data curation, simulation, and benchmarking. It includes a set of simulation-based metrics that are well-suited for evaluating sensor-based motion planning. The framework also includes a non-reactive simulation where the evaluated policy and environment do not influence each other. This allows for open-loop metric computation while being better aligned with closed-loop evaluations than traditional displacement errors. NAVSIM's evaluation protocol is better aligned with closed-loop driving, and it benchmarks an established set of end-to-end planning baselines. The results show that simple models can match recent large-scale end-to-end driving architectures in challenging scenarios. The framework also includes a challenge where 143 teams from 13 countries developed diverse methods that competed on the proposed benchmark. The top methods ranged from multi-billion parameter vision language models to more efficient and recently overlooked approaches based on trajectory sampling and scoring. NAVSIM provides a standardized evaluation server on the open-source HuggingFace platform, which is free, has low maintenance overhead, and enables future scaling to more challenging datasets and metrics. The framework is designed to address the shortcomings of existing driving benchmarks and provide standardized but configurable simulation-based metrics for benchmarking driving policies. It improves accessibility for conducting AV research by providing downloadable challenging scenario splits and simple data curation methods. The results show that NAVSIM can serve as an accessible toolkit for AV researchers that bridges the gap between simulated and real-world driving. The framework is designed to support more datasets in the future and advocate for more open dataset releases by the community for accelerating progress in autonomous driving.