Local AI now runs on more hardware, you can buy in one click, and smarter defaults mean less setup. Here’s everything in this release.
What’s new
- Local AI now runs on AMD and Intel GPUs — not just NVIDIA. Valence detects your graphics hardware and picks the right acceleration — NVIDIA via CUDA, AMD and Intel via Vulkan — and falls back to your CPU when there’s no compatible GPU. It also tells you which one it’s using, so nothing’s a mystery.
- Buy Valence in one click — purchase a license directly from the app, no detour through the website.
- Smarter defaults, less setup — local memory and file chunking are on out of the box, and Valence opens as a normal, right-sized window (no longer always-on-top).
- Your model’s ready when setup finishes — the local model you download during setup is now selected automatically.
- You won’t miss an update — when a new version is ready, Valence asks you front-and-center whether to download it now.
Fixes
- Local-model errors are clearer — a simple, kid-friendly message in Kids Mode, and a clear, actionable one everywhere else (no more wall of technical text).
- Auto-switch triggers update live when you toggle them, instead of staying greyed out until you reopen Settings.
- Your memory count updates the moment a memory is saved.
- Fixed in-app updates failing to install with “the installation package could not be opened.”
Updating from 1.0.2 or earlier? The old in-app updater can’t install this fix on its own — please update to 1.0.3 manually just this once. After that, in-app updates work normally.
Download
Valence 1.0.3 (.msi) · Windows 10/11 (64-bit) · 829 MB · 7-day free trial, then $59.95 once.
SHA-256:
9722ED5FB2DE27A5F78E82AE28A6E2F9C1107BFA3BAFCCAF6CA0329C2D1BB599Unsigned installer — SmartScreen may warn: More info → Run anyway. Existing installs upgrade in place.