Qoru Chat Beta: run open models locally, no API keys required
Qoru Chat is now in open beta — a privacy-first chat interface that runs quantized open models on your own hardware.
We built Qoru Chat because running an open-weight model locally shouldn't require a PhD in CUDA debugging. Today we're opening it up as an open beta.
What it is
Qoru Chat is a desktop interface (macOS, Windows, Linux) for chatting with open models on your own hardware. It handles model download, quantization, and GPU/CPU placement automatically — no API keys, no vendor lock-in, no internet required after initial setup.
The beta ships with out-of-the-box support for our published multilingual models and popular community checkpoints (Phi-3, Gemma, Qwen), all running through a clean, distraction-free chat UI.
What works today
- One-click model install from Hugging Face. Qoru Chat picks a quantized variant automatically based on your hardware and downloads it in the background.
- Local-first indexing for document Q&A. Drop a folder of PDFs or text files and the model will ground its answers in your documents — entirely offline.
- Low-RAM mode. For laptops that max out at 16 GB, we swap non-attention layers to disk. Expect 2–3× slower output, but it runs.
What's coming before stable
- Custom model support — bring your own GGUF / Safetensors weights.
- Multi-conversation threads with named saves.
- Plugin host for local tools (calculator, calendar, etc.).
Try it
Download the beta from github.com/qoru-gbl/qoru-chat/releases. It's free while in beta; we'll set pricing when we ship stable later this year.
Built by Paul and the Qoru engineering team. Questions? File an issue or ping us on Twitter.