# How Do Model Export Formats Impact the Development of ML-Enabled Systems? A Case Study on Model Integration

arXiv: https://arxiv.org/abs/2502.00429 · HTML: https://mlsanity.com/paper/2502.00429/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 12 of its 46 references that are in the map.
- First posted on arXiv 1 Feb 2025.

## Builds on (most cited first, 12 of 12)

- [TensorFlow-Serving: Flexible, High-Performance ML Serving](https://mlsanity.com/paper/1712.06139/) (2017)
- [Software Engineering for AI-Based Systems: A Survey](https://mlsanity.com/paper/2105.01984/) (2021)
- [Engineering AI Systems: A Research Agenda](https://mlsanity.com/paper/2001.07522/) (2020)
- [A Software Engineering Perspective on Engineering Machine Learning Systems: State of the Art and Challenges](https://mlsanity.com/paper/2012.07919/) (2020)
- [Characterizing Technical Debt and Antipatterns in AI-Based Systems: A Systematic Mapping Study](https://mlsanity.com/paper/2103.09783/) (2021)
- [Characterizing and Detecting Mismatch in Machine-Learning-Enabled Systems](https://mlsanity.com/paper/2103.14101/) (2021)
- [Adapting Software Architectures to Machine Learning Challenges](https://mlsanity.com/paper/2105.12422/) (2021)
- [Challenges and Practices of Deep Learning Model Reengineering: A Case Study on Computer Vision](https://mlsanity.com/paper/2303.07476/) (2023)
- [Green AI: A Preliminary Empirical Study on Energy Consumption in DL Models Across Different Runtime Infrastructures](https://mlsanity.com/paper/2402.13640/) (2024)
- [The Product Beyond the Model -- An Empirical Study of Repositories of Open-Source ML Products](https://mlsanity.com/paper/2308.04328/) (2023)
- [Reusing Deep Learning Models: Challenges and Directions in Software Engineering](https://mlsanity.com/paper/2404.16688/) (2024)
- [A Large-Scale Study of Model Integration in ML-Enabled Software Systems](https://mlsanity.com/paper/2408.06226/) (2024)

## Built on by (most cited first, 1 of 1)

- [Architecturally Significant MLOps Guidelines for ML Model Integration and Deployment: a Gray Literature Review](https://mlsanity.com/paper/2606.06535/) (2026)

## Abstract (arXiv)

Machine learning (ML) models are often integrated into ML-enabled systems to provide software functionality that would otherwise be impossible. This integration requires the selection of an appropriate ML model export format, for which many options are available. These formats are crucial for ensuring a seamless integration, and choosing a suboptimal one can negatively impact system development. However, little evidence is available to guide practitioners during the export format selection. We therefore evaluated various model export formats regarding their impact on the development of ML-enabled systems from an integration perspective. Based on the results of a preliminary questionnaire survey (n=17), we designed an extensive embedded case study with two ML-enabled systems in three versions with different technologies. We then analyzed the effect of five popular export formats, namely ONNX, Pickle, TensorFlow's SavedModel, PyTorch's TorchScript, and Joblib. In total, we studied 30 units of analysis (2 systems x 3 tech stacks x 5 formats) and collected data via structured field notes. The holistic qualitative analysis of the results indicated that ONNX offered the most efficient integration and portability across most cases. SavedModel and TorchScript were very convenient to use in Python-based systems, but otherwise required workarounds (TorchScript more than SavedModel). SavedModel also allowed the easy incorporation of preprocessing logic into a single file, which made it scalable for complex deep learning use cases. Pickle and Joblib were the most challenging to integrate, even in Python-based systems. Regarding technical support, all model export formats had strong technical documentation and strong community support across platforms such as Stack Overflow and Reddit. Practitioners can use our findings to inform the selection of ML export formats suited to their context.