Does Saient need an account or licence key?

No. Saient is free to download and local inference does not require a Saient account, licence key or hosted-inference API key.

Saient says my GPU isn't being used.

Saient ships a CUDA engine for NVIDIA GPUs and falls back to CPU if no GPU is detected. Make sure your NVIDIA driver is installed and nvidia-smi works. The title bar shows live GPU stats when a model is loaded.

Where do I put model files?

Saient scans its models folder for GGUF files. Drop your .gguf models there (or use the in-app picker) and they appear in the model list.

Help & Setup — Saient

Install

The current free no-key release has a verified Debian/Ubuntu amd64 package and a Windows x64 test installer. Download either directly from the download section.

Debian/Ubuntu — download Saient_1.0.9_amd64.deb, then open it with your package installer or run sudo apt install ./Saient_1.0.9_amd64.deb.
Windows x64 — download Saient_1.0.9_x64-setup.exe, open it and follow the current-user installer. This test build bundles the CPU GGUF engine. It is not yet Authenticode-signed, so Windows may identify the publisher as unknown.
AppImage — a replacement no-key build is not currently published. Do not use the obsolete v1.0.3 installers if you want the free build.

Updates after this install: turn on Settings → Internet, check for an update, then choose Install update and restart. You can turn Internet access off again afterward. Windows uses Tauri's signed current-user updater; Linux verifies Saient's signed manifest and SHA-256 before using the system package installer. Users coming from v1.0.3 must install v1.0.9 manually once because that older app did not contain these installers.

The audited Linux package bundles CPU and CUDA GGUF engines. The current Windows test package bundles the CPU GGUF engine only; Windows GPU acceleration and the optional Python studios have not yet passed an end-to-end Windows check.

First run & setup

The first time you open Saient, a setup wizard walks you through getting ready:

Fast setup — installs the essentials and bundled AI engine. You still need to choose or download a compatible local model.
Full setup — also installs the optional Python tools (in a managed virtual environment) for image generation, the vision analyzer, text-to-speech and LoRA training.

You can re-run setup or install extras later from inside the app.

Loading models

Saient uses GGUF models. Place your .gguf files in Saient’s models folder (shown in the app) and they’ll appear in the model list — pick one and click load.

The bundled GGUF engine recognises Llama, Mistral, Phi, Qwen2, Gemma, Falcon and gpt-oss architectures. Supported tensor types include Q4_K, Q5_0/Q5_1, Q6_K, Q8_0, BF16, F16 and F32.

Bigger models need more VRAM. If a model fails to load, try a smaller quantisation (e.g. a Q4_K build) or a smaller model.

GPU & CUDA

Saient ships with a prebuilt CUDA engine and its CUDA runtime library. You still need a working NVIDIA driver. On launch it detects a usable NVIDIA GPU and otherwise falls back to CPU.

Make sure your NVIDIA driver is installed and up to date.
When a model is loaded, the title bar shows live GPU stats so you can confirm it’s on the GPU.

The studios

Switch studios from the icon rail on the left.

Chat — talk to your local model; it renders live HTML artifacts in a side pane.
Agent — a real coding agent with a terminal and a plan→act→verify loop. Toggle it with Ctrl·Shift·K. Write mode is off by default for safety.
Image Gen — SDXL text-to-image with LoRA, schedulers and an optional face-detail pass.
Vision — ask questions about an image; the analyzer describes and answers locally.
Text-to-Speech — natural offline voices.
LoRA & Merge — train SDXL LoRAs and merge checkpoints.

Press ? any time for the full keyboard-shortcut list.

Free local inference

Saient is free to download. There is no Saient account, subscription, trial timer or licence-key screen.

Local inference does not require a hosted-inference API key. Models and optional downloads may have their own licences, terms and network requirements; check the model page before using one.

Launch password

You can set a launch password so the app asks for it on start — handy on a shared computer. Open the 🔒 button in the title bar to set, change, or remove it.

The launch password deters casual access to the app. It does not encrypt the files on your disk — treat it as a lock on the front door, not a safe.

Troubleshooting

A model won’t load or the app feels stuck

The model may be too large for your VRAM. Try a smaller quantisation or model. Stopping and reloading the model from the title bar clears the engine.

Image gen / vision / TTS aren’t available

These need the optional Python tools. Re-run Full setup to install them.

The GPU isn’t being used

Confirm your NVIDIA driver works (nvidia-smi on the command line). Without a usable GPU, Saient runs on CPU, which is slower.

I see a trial or licence prompt

You have an obsolete installer. Uninstall that copy and use the current free build from the download section. The current build does not require a Saient key.

FAQ

Is Saient free?

Yes. The current desktop build has no Saient payment, subscription, trial timer or licence key.

Does Saient send my prompts anywhere?

Local inference does not send prompts or outputs to Saient servers. Network activity can occur when you explicitly download models or updates, or enable network access for an agent action. See our privacy policy.

Can I use it offline?

Yes. Once the app, optional tools and models are installed, local inference works offline. Features that download models or use network-enabled agent tools naturally need a connection.

Which operating systems are supported?

The audited no-key package supports Debian/Ubuntu amd64. A Windows x64 CPU-LLM test installer is now published for installation and runtime testing. Windows GPU acceleration and optional Python studios remain unverified; a replacement AppImage is pending.