# Hanzo Edge — Hanzo AI > Deploy globally in milliseconds with ultra-low latency edge computing infrastructure. On-device inference # Hanzo Edge Run a model on the machine in front of you. Edge is a Rust runtime that loads a quantized model and generates tokens locally — on a laptop, an ARM board, or a browser tab. Nothing is sent anywhere, because there is nothing to send it to. Get Started View Documentation ### Where it runs macOS on Apple Silicon and Intel, Linux on x86 and ARM, and modern browsers through WebAssembly ### No network at all Weights on disk, tokens out of local memory. There is no request, so there is nothing to be slow ### Fetched once Name a Hugging Face repo and it downloads and caches the weights. After that, offline ## What the runtime does One Rust binary, a core crate to embed, a WASM build for the browser, and a local server for everything else ### A server on localhost hanzo-edge serve puts an inference endpoint on a port you pick. Point a client at it and the rest of your code does not change. ### Quantized as a first language GGUF is the native format — Q4_K, Q5_K, Q8_0. A 4B model at Q4_K_M is about 2.5GB on disk, which is what makes a phone a plausible place to run one. ### It uses what it finds Metal on Apple Silicon, CUDA where there is an NVIDIA card, AVX2 or AVX-512 on CPU. Detected at startup rather than configured by you. ### Zen models sized for devices zen3-nano at 600M for embedded work, zen-eco at 4B for phones and tablets, zen4-mini at 8B for a laptop. All published pre-quantized. ### Token by token Streaming over server-sent events from the local server, or a callback from the Rust API, so a UI starts drawing before generation finishes. ### The prompt never leaves Nothing is uploaded and there is no key to leak. The prompt and the output stay in the process that made them, which is the whole reason to run a model here. ## When to reach for it Use Edge when the data cannot leave, the network cannot be relied on, or a per-call bill is the wrong shape. Use Hanzo Engine when you need full precision or many users at once. ### Apps that work with no signal The model is a file on disk, so the feature does not stop existing when the connection does. - •A desktop app that drafts and summarizes without a round trip - •Field tools on a laptop, on a site with no coverage - •Anything that has to keep working on a plane ### Text that cannot leave the building Some prompts are a legal problem the moment they cross a network boundary. This is the answer to those. - •Clinical and legal text processed on the workstation that holds it - •On-premise deployments with no egress at all - •Personal data handled on the device it belongs to ### Inside the browser tab The WebAssembly build runs the model in the page, with nothing behind it — so a visitor costs you bandwidth once and compute never. - •A demo that does not bill per visitor - •An assistant inside a web app you already ship - •Classification done client-side, before anything is uploaded ### Small hardware ARM64 Linux is a production target. Cortex-A class boards are experimental, and the 600M model is the one that fits them. - •A 600M model on a single-board computer - •An ARM server doing batch work on its own - •Kiosks and appliances with no account to sign in to ## Install it, name a model cargo install hanzo-edge, then hanzo-edge run --model zenlm/zen3-nano. It downloads the weights the first time and never needs the network again. Apache-2.0. Install Read the docs ## Get started with Edge [Read the docs](https://docs.hanzo.ai/docs/edge)[View on GitHub](https://github.com/hanzoai) ## Open source License: Apache-2.0[hanzoai](https://github.com/hanzoai) ## Get Edge Edge compute [Deploy to Cloud](https://console.hanzo.ai/deploy)[Self-host](https://docs.hanzo.ai/docs/edge)