Valence

A private, controlled, curated AI space — every AI model you use, local or cloud, on your terms.

Website · Updates · All releases · help@helixailabs.com

⬇ Download

Windows 10/11 (64-bit) · 829 MB · 7-day free trial, then $59.95 once · works fully offline · no subscription

SHA-256: C58BCFCE31C74BD76DC8CC7DD8F73587460C385C4A9DBC2A18BC94779BCC0B4D The installer is unsigned, so Windows SmartScreen may warn you — click More info → Run anyway. Existing installs upgrade in place.

Why is the download ~829 MB?

Valence ships complete, offline GPU acceleration in the box - no separate CUDA toolkit, no multi-gigabyte driver detour, no account. Most of the size is NVIDIA’s GPU math libraries:

Piece	Approx. size	What it does
NVIDIA CUDA libraries (cuBLAS / ggml-cuda)	~640 MB	Fast local AI on NVIDIA GPUs
Valence app + bundled .NET runtime	~190 MB	The app itself, self-contained (no .NET install needed)
Vulkan backend	~54 MB	Local AI on AMD & Intel GPUs
Embedding model + bundled tools	~45 MB	On-device memory, knowledge, and web search

Everything - your conversations, models, and memory - runs on your machine. The size is the cost of an AI that works fully offline and never phones home.

Getting lighter soon. Valence already detects your hardware and uses the right acceleration (NVIDIA CUDA, AMD/Intel Vulkan, or CPU). A coming update will download only the pack your machine needs during setup, bringing the base install down to around 190 MB.

What is Valence?

A private, controlled, curated space for working with AI — local or cloud, all under your control.

Curated, cross-model — OpenAI, Anthropic, Google, Azure, xAI, OpenRouter, Ollama, and on-device local models, all from one interface.
Private by design — conversations, memory, and documents stay on your computer. No account required.
Fully local option — download a model (Gemma, Qwen, …) and chat with nothing leaving your PC.
You’re in control — local memory you own, per-profile knowledge from your own documents, model compare, and X-Ray (“why did it answer that?”).
Kids Mode — a locked-down chat for children with a parent dashboard, on-device safety, and daily time limits.

Releases

Version	Date	Summary
1.0.5 (latest)	Jun 27, 2026	Smoother updates: the startup update prompt now appears, a robot animation shows the download, and old installers are cleaned up automatically.
1.0.4	Jun 27, 2026	Lens and X-Ray reimagined as clean anchored windows over the canvas, plus a custom-lens authoring fix.
1.0.3	Jun 26, 2026	Local AI on AMD & Intel GPUs (Vulkan) with CPU fallback, one-click in-app purchase, smarter defaults, and an in-app update fix.
1.0.2	Jun 26, 2026	Interface polish, friendlier first-run, Ollama installed-model listing, manual Check for Updates, NVIDIA driver guidance.
1.0.1	Jun 23, 2026	Installer packaging fix — bundles the complete app asset set.
1.0.0	Jun 22, 2026	First public release.

Full notes for every version live on Updates.