executorch
On-device AI across mobile, embedded and edge for PyTorch
Decision gist · record as of 2026-08-14
Yes, if you need to deploy PyTorch models to mobile or edge devices. The active maintenance, permissive license, and platform-specific prebuilt backends make it a practical choice for on-device AI. Install with conditions: verify that your model uses only core ATen operators, confirm backend availability for your target platform, and plan for the 21 runtime dependencies unless you use the minimal build mode.AI-flagged interpretation of the facts on this page — verify before relying
Before you install
- Requires torch to be installed separately; prebuilt runtime supports only core ATen operators and specific backends (XNNPACK always; Core ML and MPS on macOS; QNN on Linux x86_64; OpenVINO requires separate pip install executorch[openvino]).
- Medium install friction: 21 runtime dependencies including torch, numpy, pandas, and optional backends (coremltools, hydra-core, omegaconf).
- A minimal build mode available via EXECUTORCH_BUILD_MINIMAL=1 reduces the dependency footprint for export-only use.
License · maintenance · safety
BSD-3-Clause (permissive) — BSD-3-Clause permissive license allows commercial and private use with minimal restrictions; you must include a copy of the license and disclaimer.
last release 2026-08-14 (0 days) · last repo commit 2026-08-14 · 4,916 stars
0 known vulnerabilities (OSV.dev, 2026-08-14) · 448,392 downloads/mo, #6,601 on PyPI
Alternatives
Verify before relying
pip install executorch torch
import executorch.runtime
# Load and run a .pte model file
runtime = executorch.runtime.Runtime()
result = runtime.run(model_path)- Whether the prebuilt runtime's core ATen operator set is sufficient for your model architecture
- Performance characteristics and latency benchmarks for your target hardware
- Compatibility of your PyTorch model version with the export pipeline
What it is and what it does
ExecuTorch is a PyTorch deployment platform that bridges the gap between model development and on-device execution. It provides tools to export trained PyTorch models into a portable `.pte` format optimized for mobile, embedded, and edge hardware—including AR/VR wearables, iOS, and Android devices. The package includes both an export path (EXIR) for converting models and a prebuilt runtime module for executing those models with hardware acceleration.
The runtime is built around a restricted but efficient operator set and pluggable backend delegates. The prebuilt wheel includes XNNPACK for general acceleration, with platform-specific backends: Core ML and MPS on macOS, QNN on Linux x86_64, and OpenVINO on Linux. This architecture trades some model flexibility for predictable performance and small binary size on constrained devices. A minimal build mode is available for export-only workflows, reducing dependencies significantly.
Use it for
- Export a trained PyTorch model and deploy it to iOS or Android apps without requiring the full PyTorch runtime on the device.
- Accelerate inference on edge devices using hardware-specific backends like Core ML on Apple Silicon or QNN on Qualcomm processors.
- Run large language models locally on mobile devices by exporting and quantizing them with ExecuTorch's optimization pipeline.
- Build AR/VR applications that execute neural networks directly on wearable hardware with minimal latency and power consumption.
- Create a minimal export-only environment for model conversion without bundling runtime dependencies like pandas or hydra-core.
Worth the install?
AI-flagged interpretation of the facts on this page. Verify before relying on it.
Yes, if you need to deploy PyTorch models to mobile or edge devices.
The active maintenance, permissive license, and platform-specific prebuilt backends make it a practical choice for on-device AI. Install with conditions: verify that your model uses only core ATen operators, confirm backend availability for your target platform, and plan for the 21 runtime dependencies unless you use the minimal build mode.
Install
executorch on PyPI
Before you install
Medium install friction: 21 runtime dependencies including torch, numpy, pandas, and optional backends (coremltools, hydra-core, omegaconf). A minimal build mode available via EXECUTORCH_BUILD_MINIMAL=1 reduces the dependency footprint for export-only use. Package is actively maintained with recent releases.
Requires torch to be installed separately; prebuilt runtime supports only core ATen operators and specific backends (XNNPACK always; Core ML and MPS on macOS; QNN on Linux x86_64; OpenVINO requires separate pip install executorch[openvino]).
License in practice
BSD-3-Clause permissive license allows commercial and private use with minimal restrictions; you must include a copy of the license and disclaimer.
Quickstart
pip install executorch torch
import executorch.runtime
# Load and run a .pte model file
runtime = executorch.runtime.Runtime()
result = runtime.run(model_path)Verify before relying
- Whether the prebuilt runtime's core ATen operator set is sufficient for your model architecture
- Performance characteristics and latency benchmarks for your target hardware
- Compatibility of your PyTorch model version with the export pipeline
Package facts
| License | BSD-3-Clause permissive |
| Python support | Supports the current Python release <3.15,>=3.10 |
| Install friction | Medium. Platform-specific wheel |
| Runtime dependencies | 21 packagesexpecttestflatbuffershypothesiskgbmpmathnumpypackagingpandasparameterizedpytorch-tokenizerspyyamlruamel.yamlsympytabulatetorchtorchaotyping-extensionscoremltoolsscikit-learnhydra-coreomegaconf |
| Maintenance | Actively maintained 0 days since the last release |
| Last repo commit | |
| First released | |
| Downloads | 448,392 / month, #6,601 on PyPI 30-day window, as of 2026-08-14 |
| Known vulnerabilities | None known OSV.dev, checked 2026-08-14 |
| Classifiers | Development Status :: 5 - Production/StableIntended Audience :: DevelopersIntended Audience :: EducationIntended Audience :: Science/ResearchProgramming Language :: C++Programming Language :: Python :: 3Programming Language :: Python :: 3.10Programming Language :: Python :: 3.11Programming Language :: Python :: 3.12Programming Language :: Python :: 3.13Programming Language :: Python :: 3.14Topic :: Scientific/EngineeringTopic :: Scientific/Engineering :: Artificial IntelligenceTopic :: Scientific/Engineering :: MathematicsTopic :: Software DevelopmentTopic :: Software Development :: LibrariesTopic :: Software Development :: Libraries :: Python Modules |
Evidence: executorch-1.4.1-cp310-cp310-macosx_14_0_arm64.whl; executorch-1.4.1-cp310-cp310-manylinux_2_28_aarch64.whl; executorch-1.4.1-cp310-cp310-manylinux_2_28_x86_64.whl; executorch-1.4.1-cp310-cp310-win_amd64.whl; executorch-1.4.1-cp311-cp311-macosx_14_0_arm64.whl; executorch-1.4.1-cp311-cp311-manylinux_2_28_aarch64.whl; executorch-1.4.1-cp311-cp311-manylinux_2_28_x86_64.whl; executorch-1.4.1-cp311-cp311-win_amd64.whl; executorch-1.4.1-cp312-cp312-macosx_14_0_arm64.whl; executorch-1.4.1-cp312-cp312-manylinux_2_28_aarch64.whl; executorch-1.4.1-cp312-cp312-manylinux_2_28_x86_64.whl; executorch-1.4.1-cp312-cp312-win_amd64.whl; executorch-1.4.1-cp313-cp313-macosx_14_0_arm64.whl; executorch-1.4.1-cp313-cp313-manylinux_2_28_aarch64.whl; executorch-1.4.1-cp313-cp313-manylinux_2_28_x86_64.whl; executorch-1.4.1-cp313-cp313-win_amd64.whl; executorch-1.4.1-cp314-cp314-macosx_14_0_arm64.whl; executorch-1.4.1-cp314-cp314-manylinux_2_28_aarch64.whl; executorch-1.4.1-cp314-cp314-manylinux_2_28_x86_64.whl; executorch-1.4.1-cp314-cp314-win_amd64.whl
Tags
Let your AI agent find packages like this
Example. Real query, live index.
You found this page by searching. An agent finds it by wishing: SkillFed indexes 14,416 PyPI packages by what they can do, searchable in plain language.
wish › “hardware accelerated inference”
- executorchExecuTorch exports and runs PyTorch models on mobile, embedded, and…
- litert-lm-builderProvides Python tools for building, inspecting, and working with…
- optimumOptimum provides optimization tools to export and run Transformers,…
Give your agent the search over MCP, or paste the wish link into any chat.
More Software Development packages
Provides backported and experimental type hints for Python 3.9+, allowing use of newer typing features on older Python versions and enabling early experimentation with type system PEPs before they enter the standard library.
NumPy provides an N-dimensional array object and a comprehensive suite of mathematical, linear algebra, Fourier transform, and random number functions for scientific computing in Python.
FastAPI is a Python web framework for building REST APIs using type hints, with automatic request validation, serialization, and interactive API documentation.
Provides a way to document function parameters, class attributes, return types, and variables inline using Python's `Annotated` type hint syntax instead of traditional docstrings.
Typer builds command-line applications from Python functions using type hints, automatically generating help text, argument parsing, and shell completion.
Install it if you are building CLIs in Python.
Distlib provides low-level packaging utilities for building, distributing, and managing Python software—including metadata handling, version specifiers, wheel support, script installation, and dependency resolution.
See also tflite-runtime · ai-edge-litert · mobly · pte-adapter-model-explorer · mediapipe · optimum · litert-torch · mnn · ncnn · torch-npu