Skip to content

Onyx Code 1.8.0

Latest
Latest

Choose a tag to compare

View all tags
@onyx-app-group onyx-app-group released this 13 Jul 17:52
v1.8.0
8c3c6e7

Local AI can now run on your graphics card on Windows, with a new control to choose GPU or CPU and clearer hardware and update screens.

Downloading

If you're installing manually:

macOS

  • Apple Silicon Mac (M1 or newer): download OnyxCode-darwin-arm64.dmg (smaller, native).
  • Intel Mac, or unsure: download OnyxCode-darwin-universal.dmg (runs on both).

Open the .dmg and drag Onyx Code to /Applications/. Both macOS downloads are signed and notarized by Apple, so they pass Gatekeeper and open without security warnings.

Linux (x64)

  • Debian / Ubuntu: download OnyxCode-linux-x64.deb and install it with sudo apt install ./OnyxCode-linux-x64.deb.
  • Other distributions: download OnyxCode-linux-x64.AppImage, make it executable (chmod +x), and run it. On Ubuntu 24.04 (which lacks libfuse2) run it with --appimage-extract-and-run.

Windows (x64)

Windows support is on the way through winget (winget install OnyxCode), which will be the recommended way to install once it's published: it runs through Microsoft's App Installer (a trusted Microsoft Store app) with a package listing reviewed in Microsoft's community winget repository, so it verifies the download and gives a smoother first-run experience. winget is not available yet, so for now the only way to install on Windows is to download the attached OnyxCode-win32-x64-winget.exe and run it manually. It is not code-signed yet, so Windows SmartScreen warns "Windows protected your PC"; click More info, then Run anyway to continue. A code-signed installer is planned for a later release.

The *-autoupdate.zip, darwin-*.json, linux-x64.json, and win32-x64.json files are plumbing for Onyx Code's
built-in auto-updater. They are not designed for manual install and will give
you a worse first-run experience than the download above.

What's new

GPU acceleration for local AI on Windows. Onyx Code can now run local models on your graphics card on Windows, so chat and autocomplete are much faster than running on the processor. It works with NVIDIA, AMD, and Intel graphics, with no setup and no extra downloads.

Choose where AI runs: Auto, GPU, or CPU. A new processor control lets you run local AI on your graphics card or your processor. Set it in the "Your Hardware" onboarding step, from the Command Palette with "Onyx: Select AI Processor", or in chat with /processor. Auto, the default, uses the graphics card wherever it can and the processor otherwise.

A clearer hardware screen. Onboarding now shows exactly where local AI runs on your machine, such as "GPU (Vulkan)", "GPU (Metal)", or "CPU" with the reason, so you always know what is powering your local models.

Update downloads you can see. Downloading an update now shows a progress notification while it runs, with a size readout, so you can tell it is working at a glance. When it finishes, Restart to Update switches you over with your work and settings preserved.

System requirements

  • macOS 11 (Big Sur) or later, on Apple Silicon or Intel.
  • Linux: a 64-bit (x64) distribution; Debian/Ubuntu via the .deb, other distributions via the .AppImage.
  • Windows 10 or later (x64).

Verification

This build is signed and notarized by Apple. macOS verifies it on first launch, so it opens as an app from an identified developer without security warnings.

Assets 13
Loading