AI Potluck
Contribute
Back to Gap Map Model components / Inference code

ExecuTorch

PyTorch Foundation (Linux Foundation)
open source / Overall score: 4.5(leading)

ExecuTorch is PyTorch's stack for running models on device, from phones, laptops and browsers down to microcontrollers. Models captured with torch.export are lowered to a .pte file for each target and run by a C++ runtime that links only the operators a deployment needs, with backends from Apple, Arm, Qualcomm, MediaTek, Samsung, Intel, NVIDIA and others.

Openness

5 medium confidence
5.0
license
BSD-3-Clause(OSI)
source
public(the published repository is the engine)
core features withheld
no — no enterprise path in the repository root and no paid build of the engine in the README

ExecuTorch's LICENSE is the stock BSD-3-Clause text, with no added restriction, and the public repository is the engine itself rather than a wrapper around closed code. Its root carries no enterprise or commercial directory, and the README describes no license-gated build, so nothing in the published source is held back.

Adoption

5 medium confidence
5.0

ExecuTorch ships inside Meta's own apps and devices -- Instagram, WhatsApp, Messenger, Quest and Ray-Ban Meta smart glasses -- reaching billions of people. GitHub stars measure attention to a repository rather than a runtime bundled invisibly inside other companies' apps, so adoption is read from that deployment breadth instead.

Capability

4 medium confidence
4.0

ExecuTorch is a tier below vLLM, level with ONNX Runtime, judged by execution-provider breadth across CPU, GPU, mobile and edge hardware. Its backend table covers more silicon vendors than Apple Core ML Runtime's, from phones and desktops down to microcontrollers, but it executes models on device rather than serving them with batching, parallelism and a multi-tenant API surface.

Verified 2026-10-07