Valence 1.4.1 - Context that fits your PC, clearer model switching
Two improvements to running AI locally on your own machine - still private, still on your hardware.
What’s new
- Your local AI fits the conversation to your hardware. Until now an on-device model kept only a short, fixed slice of the chat in view - so it started forgetting earlier messages just a few pages in, even on a capable PC. Valence now sizes that window to your actual hardware and the model you’re running: it follows much longer conversations where your machine allows, and stays lean where it doesn’t - so Valence is never the cause of a system slowdown. Leave it on Auto (recommended) or pick a size yourself under Settings → Local Model. If you’re still on the old fixed setting, Valence offers to switch you when it can do better.
- Clearer model switching. When you switch a local model, Valence now tells you what’s actually happening - the model’s name, the current step (loading the model, warming up), and a live timer - and starts loading it right away, instead of a vague “getting ready” that left you guessing.
Download
Two installers - both ship every release:
Valence 1.4.1 - All-in-one (.msi) · 809 MB · CUDA + Vulkan bundled, fully offline.
Valence 1.4.1 - Slim (.msi) · 303 MB · your GPU pack is fetched on first run.
Windows 10/11 (64-bit) · 7-day free trial, then $59.95 once.
All-in-one SHA-256:
65533F208D11DB220DBF13080845DCEA139FF4F43085DB8C34FCE29A816C7EC2Slim SHA-256:0166EF6F678998E935888007A853B2392FB7CC2D0F6B006115979AD2E9C648F8Unsigned installer - SmartScreen may warn: More info → Run anyway. Existing installs upgrade in place.