EdgeLM: On-device AI Runtime

EdgeLM is a shared, on-device AI runtime for Android. It runs language models
directly on your phone — privately, offline, and shared across every app that wants
to use AI. Think of it as a system service for local intelligence: download a model
once, and any EdgeLM-powered app can use it, sharing a single copy in memory instead
of each app bundling its own.

PRIVATE BY DESIGN
Your prompts and the model's responses never leave your device. There is no cloud,
no account, no tracking, and no ads. Inference happens locally, in memory, and is
never stored or transmitted.

WORKS OFFLINE
Once you've downloaded a model, EdgeLM runs with no internet connection at all. It's
the same on a plane, in a tunnel, or on airplane mode.

SIMPLE TO SET UP
A quick welcome explains how it works, then EdgeLM suggests the right model for your
phone — one tap to download and you're chatting in the built-in playground. Prefer
control? Switch to Advanced any time for the full catalog with sizes, context lengths
and licenses.

CHOOSE YOUR MODEL
Pick from a curated catalog spanning tiny-and-fast to more capable:
• Qwen2.5 0.5B — runs on almost any phone
• Llama 3.2 1B — great everyday balance, 128K context
• Qwen2.5 1.5B — sharper reasoning, multilingual
• Llama 3.2 3B — higher-quality writing and reasoning
• Phi-3.5 mini — strong at code and math (flagship phones)

Each model shows its size, memory needs, license, and what it's best for — with a
warning if your device may not have enough RAM. Keep several installed and switch
between them instantly.

SHARED, EFFICIENT RUNTIME
EdgeLM loads a model once and serves every app on your device from that single copy,
so a second app adds almost no extra memory. A live notification shows what the
runtime is doing, lets you free memory on demand, and the runtime automatically
releases memory when idle.

BUILT-IN PLAYGROUND
A simple chat screen lets you test the runtime and the active model right away —
stream a reply, see tokens-per-second, stop mid-generation, and start fresh. Great
for confirming everything works before other apps use it.

FOR DEVELOPERS
EdgeLM exposes a tiny, OpenAI-style interface so your app can call on-device AI in a
few lines, without shipping or managing model weights yourself. Apps request access
through a dedicated permission, and the runtime serves them by priority.

WHY ON-DEVICE?
• Privacy — data stays on the phone
• Offline — no connection required
• Cost — no per-request cloud fees
• Speed — no network round-trips
• Efficiency — one shared model for all apps

EdgeLM is open about what it does: no ads, no tracking, no cloud inference. Just fast,
private AI that lives on your device.