README.md

---
title: Edge0
emoji: ⚡
colorFrom: indigo
colorTo: purple
sdk: static
pinned: false
---

<div align="center">

<img src="./banner.jpg" alt="Edge0 — On-Device Native Foundation Models" width="100%">

<br>

**On-device AI built for consumer hardware**<br>
Cloud-level AI · Offline access · Strengthened privacy · No pay-per-use fees

[Website](https://edge0.ai/) · [Models](https://huggingface.co/Edge0) · [GitHub](https://github.com/Edge0-AI) · [X](https://x.com/Edge0_AI)

</div>

Edge0 builds efficient foundation models for phones, laptops, wearables, robots, and other consumer hardware. Our models are designed for local, private, and offline inference.

## LLMs

- [**Edge0-35B-A3B**](https://huggingface.co/Edge0/Edge0-35b-a3b-preview) — 35B sparse MoE with about 3 GB peak active memory.
- [**Edge0-8B-A1B**](https://huggingface.co/Edge0/Edge0-8B-A1B-preview) — lightweight 8B sparse MoE with about 1 GB peak active memory.

## Speech models

- **Audio8-ASR** — [3B](https://huggingface.co/Edge0/ARK-ASR-3B) · [0.6B](https://huggingface.co/Edge0/ARK-ASR-0.6B) · [0.1B](https://huggingface.co/Edge0/Audio8-ASR-0.1B) — speech recognition from flagship to edge tiers.
- **Audio8-TTS** — [0.6B](https://huggingface.co/Edge0/Audio8-TTS-Preview-0.6b) · [0.1B](https://huggingface.co/Edge0/Audio8-TTS-Preview-0.1b) — multilingual, zero-shot voice generation with voice cloning.

Explore all models on [Hugging Face](https://huggingface.co/Edge0), and the open-source inference stack on [GitHub](https://github.com/Edge0-AI).

[Research](https://edge0.ai/research) · [About](https://edge0.ai/aboutus) · [Contact](mailto:dev@edge0.ai)

<div align="center">

**AI that consumer hardware can actually run.**

</div>