# Burn

> MIT/Apache-2.0 Rust deep-learning framework — train and run on many backends

**Category:** Rust  
**Pricing:** open-source  
**URL:** https://burn.dev  
**GitHub:** https://github.com/tracel-ai/burn  
**Added:** 2026-10-09

---

## The Hook

Burn is a deep-learning framework written in Rust for both training and inference, with swappable backends so the same model code runs on CUDA, Metal, Vulkan/WebGPU, CPU, or in the browser via WASM.

## What's Good

Dual-licensed MIT OR Apache-2.0 (tracel-ai/burn), v0.22.0. Tensor API with autodiff, kernel fusion, and backend choice at compile time; ONNX import to bring in models trained elsewhere; ships a single Rust binary with no Python runtime. Useful for embedding inference in Rust services, edge devices, or WASM. ★~16k.

## The Catch

Far fewer ready-made models, tutorials, and pretrained weights than PyTorch — ONNX import helps, but not every operator converts. Pre-1.0: expect breaking API changes between releases and pin your version. Smaller community for debugging training issues.

## Verdict

MIT/Apache Rust DL framework when you want training or inference without Python — accept a thinner model zoo and pre-1.0 churn.