Local hardware-accelerated AI desktop assistant (Vulkan GPU offload build for AMD/Intel GPUs)
Qube is a fully local, privacy-first, voice-to-voice AI desktop assistant. It integrates speech-to-text, text-to-speech, retrieval-augmented generation, and local LLM inference into a native PyQt6 desktop shell. This package is the Vulkan GPU offload build for AMD and Intel GPUs. Install only one Qube variant; user data is shared across variants.
| Publisher | dagaza |
|---|---|
| Version | 1.3.53 |
| License | MIT |
| Installer type | inno |
| Architecture | x64 |
| Package ID | dagaza.Qube.Vulkan |
| Dependencies | KhronosGroup.VulkanRT |
winget install --id=dagaza.Qube.Vulkan -ewinget upgrade --id=dagaza.Qube.Vulkan -ewinget uninstall --id=dagaza.Qube.Vulkan -e"Qube-1.3.53-vulkan-Setup.exe" /VERYSILENT /SUPPRESSMSGBOXES /NORESTARTYes. Qube (Vulkan) is free and open-source software released under the MIT license.
The latest version is 1.3.53.
The download button links to the official installer published on github.com, the same file Windows Package Manager installs. You can verify the downloaded file by comparing its SHA-256 checksum with 012813ca70fd718f0486a0baf8f547d08e712b96d5147889328d71efddfa30d3.
Qube (Vulkan) runs on Windows and is available for 64-bit (x64) systems.
Run the offline installer with "Qube-1.3.53-vulkan-Setup.exe" /VERYSILENT /SUPPRESSMSGBOXES /NORESTART, or install it with Windows Package Manager: winget install --id=dagaza.Qube.Vulkan -e --silent.
Open Settings › Apps › Installed apps, find Qube (Vulkan) and select Uninstall. From the command line, run winget uninstall --id=dagaza.Qube.Vulkan -e.
Other AI for Productivity programs for Windows from different developers.
Local-first personal AI assistant — your data stays on your machine.
Private, local AI desktop app — run open LLMs fully offline with a coding agent, knowledge base and voice.
A coding agent that lives on your computer — and can actually run it.
Sovereign AI agents that run 100% on your own machine.
AI that gets your everyday tasks done.
A team of long-lived bots for DeepSeek Harness.
