Quo
Quo is a free, open-source private AI chat app under Apache-2.0. It runs 11 open-weight models from Qwen, ranging from a 0.8B-parameter text model at 0.5 GB to a 27B at 17.4 GB, on the user's own device. The web version uses WebAssembly, desktop uses CUDA on NVIDIA GPUs through an Electron app, and iPhone uses Metal. One caveat: iPhone and iPad support is listed as upcoming at the time it was read.
What does Quo do?
Quo describes itself as "Private AI that works offline." Chat, image analysis and document understanding happen through an on-device model, with voice input and text-to-speech on top. The reasoning for the thinking models is readable in the UI, and memory is held on the device rather than in a cloud account.
Three code paths back the three form factors. The web version loads the model through WebAssembly, so a browser alone is enough. The desktop app is Electron and uses CUDA on NVIDIA GPUs where present, falling back to CPU. The mobile app is React Native and uses Metal on iPhone and Core ML. An optional cloud model path is marked clearly in the UI so a user knows when the request leaves the device.
Key facts
- Licence: Apache-2.0.
- Platforms: web (Chrome and Edge, WebAssembly); desktop (Electron on macOS, Windows, Linux x64 and ARM); mobile (React Native on Android, with iPhone and iPad listed as upcoming).
- Models: 11 open models from Qwen, between 0.8B parameters at 0.5 GB and 27B at 17.4 GB. Reasoning ("thinking") and vision models are included.
- Features: document understanding, photo analysis, voice input, text-to-speech, device-local memory, code mode with Git integration on desktop, web search.
- Model locations: on-device by default. Optional cloud models are clearly labelled.
- Hardware acceleration: CUDA on NVIDIA GPUs (desktop), Metal on iPhone (planned), CPU everywhere else.
- Install: download the desktop app from the Releases page; install the Android app from the Play Store or APK; open the web app in Chrome or Edge.
What does it replace, and where does it fall short?
Quo replaces ChatGPT for users who want a private chat UI, with the model on the device by default, across web, desktop and mobile. It fits where a reader wants one product that runs the same chat on three form factors with open-weights models.
Where it falls short: iPhone and iPad are listed as upcoming, so Apple mobile users cannot install Quo yet at the time it was read. The models are limited to Qwen's open-weights family (0.8B to 27B), so no Llama, Mistral or DeepSeek weights come with the app. On a laptop without a GPU, only the smaller models are practical; the 27B model needs 17.4 GB of RAM and GPU acceleration to be responsive. The optional cloud model path is tied to the maintainers' choice, not any provider.
How does Quo run?
Install the desktop app from the Releases page on macOS, Windows or Linux. Install the Android app from the Play Store or install the APK by hand. Open the web app in Chrome or Edge. On first launch the app downloads a model from the 11-strong catalogue; smaller models start at 0.5 GB and can be picked for lower-end hardware.
A chat turn runs on the device. The desktop app uses CUDA when an NVIDIA GPU is present, Metal on an Apple Silicon Mac, and CPU otherwise. Document understanding and photo analysis use the vision-capable models in the catalogue. The web build uses WebAssembly and runs in a tab, with the model file cached in the browser.
Who is Quo for?
A reader who wants one private chat app across a work laptop, a phone and a browser, and is happy with Qwen's open-weights models. A reader who wants Llama or Mistral weights should look at Open WebUI paired with Ollama or llama.cpp. A reader on an iPhone or iPad should wait, or look at a hosted on-device app.
What limits does the README state?
- iPhone and iPad are listed as upcoming at the time it was read.
- Only Chrome and Edge are listed for the web app; Safari and Firefox are not named.
- Larger models need GPU acceleration to be responsive on a chat turn.
- The 11 models in the catalogue are from Qwen; other open-weights families are not bundled.
- Cloud models are an opt-in, clearly labelled in the UI.
Questions people ask
Is Quo open source?
Yes. The repository is under Apache-2.0, an OSI-approved licence. The Qwen open-weights models the app downloads carry their own model licence, published by Alibaba's Qwen team on Hugging Face. The README calls Quo "free and open source" and says device inference is the default path.
Does Quo work offline?
Yes. After the first-run model download, chat turns run on the device, and the README's "Private AI that works offline" line is the project's own framing. The optional cloud model path is labelled in the UI so a user knows when a reply is sent to a server; it is off unless a cloud model is picked explicitly.
Can Quo be used commercially?
Yes for the code, which is Apache-2.0. The Qwen model licences on Hugging Face have their own commercial terms: most Qwen open-weights models allow commercial use below a user count, with a separate licence above that. A commercial user should read the model card for the exact model they ship with, because the terms differ by size.
How does Quo compare with ChatGPT?
ChatGPT runs on OpenAI's servers, keeps user data per the ChatGPT terms, and ships a single commercial product. Quo runs the model on the device by default, keeps memory on the device, and is Apache-2.0 open source across web, desktop and Android. The trade is: ChatGPT's hosted model quality and polish against on-device privacy. The open-source alternatives to ChatGPT page sets Quo beside Open WebUI, Jan and LibreChat.
Which Qwen model should a laptop user pick?
The README names 11 models between 0.8B and 27B. The 0.8B model at 0.5 GB runs on most laptops without a GPU and is useful for quick tasks. The 7B-class models give better reasoning on machines with 16 GB of RAM. The 27B model at 17.4 GB needs a GPU with enough memory or a workstation; otherwise response times are slow. The README does not publish measured latency figures per chip.
Sources
- Quo README and LICENSE: github.com/quo-ai/quo, read 2026-11-09.
More chat interfaces
A self-hosted chat platform for Ollama and OpenAI-compatible models with document search over 9 vector databases, web search, image generation, roles and single sign-on, under a BSD-3 licence with an added branding clause.
LobeHub, formerly LobeChat, is an agent workspace with an agent builder, agent groups, scheduled runs, editable memory and a library of over 10,000 tools and MCP plugins, under a community licence that needs a paid licence to distribute a modified version.
An MIT all-in-one AI app from Mintplex Labs for chatting with your documents with citations, running agents with MCP, and using any of 39 model providers including local llama.cpp and Ollama, as a desktop app or a multi-user Docker server.
An AGPL-3.0 desktop client for Windows, macOS and Linux that connects OpenAI, Gemini, Anthropic and local Ollama or LM Studio models, with 300+ preset assistants, side-by-side model chats, file support and MCP servers.
An AGPL-3.0 desktop app for local models, formerly text-generation-webui, with five engines including llama.cpp and ExLlamaV3, an OpenAI and Anthropic-compatible API, tool calling, file chat, image generation and LoRA training, fully offline.
Added November 9, 2026. Every claim above comes from the project's README, LICENSE or model card, read on November 9, 2026, or from the GitHub API on the date shown in the panel. Found an error? Write to reveneau@licheo.com and it is fixed in the next weekly pass. Repository: github.com/quo-ai/quo.
Want Quo running on your own servers?
Reveneau sets it up on your hardware, connects it to your product, and builds the software around it.