Your library, your way
Download recommendations, keep local files, or link existing artifacts without making another copy. Give each model an editable API alias.
NINFEREZ MANAGER
Local AI.
All together.
A convenient Windows workspace built around NInfer. Bring model libraries, GPU engines, inference profiles, and API access into one place.
Manager workspace · Windows x64 · MIT licensed
NInferEZ ManagerYOUR WORKSPACE
Local firstEverything in one place.
Ready when you are.
Model libraryLocal files & linked models
GPU engineA package for your GPU
Your models NInfer v3
S
Swift 1.527B · NVFP4 · Text + Vision
VisionO
OrcaRouter27B · IQ3_XXS · MTP + DFlash2
Text127.0.0.1:8173/v1YOUR LOCAL APIA LIBRARY THAT IS YOURS
Your models. Organized.
Download, import, or link an existing file.
S
Swift 1.5NVFP4 · 21.22 GiB artifact
VisionO
OrcaRouterIQ3_XXS · 10.06 GiB artifact
TextO
OrcaRouterIQ3_XXS · 12.13 GiB artifact
TextManaged profiles help you choose a starting point for your model and available VRAM.
FROM SOURCE TO ARTIFACT
A guided conversion.
The same backend as the standalone Converter.
- 1Choose your sourceLocal checkpoint or supported GGUF
- 2Check compatibilityResources, geometry, and prepared plans
- 3Review & convertNative NInfer artifact, with progress
THE RIGHT ENGINE
Matched to your hardware.
Choose a package for your NVIDIA GPU.
RTX 3000Compatible architecture targets
sm86RTX 4000Compatible architecture targets
sm89RTX 5000Compatible architecture targets
sm120aSHA-256 integrity checks before activation.
Explore the workspace Interactive product illustration
BUILT AROUND YOUR WORKFLOW
One workspace.
The control you need.
Profiles with a purpose
Managed settings coordinate dependent controls. Advanced manual settings stay available, with clear compatibility errors.
Engine management
GPU-specific packages, SHA-256 verification, activation, and update checks. Models and engines are managed independently.
An API that is ready
A local API starts with Manager. Load models on request, cancel loading, or adjust idle unloading to reclaim GPU memory.
Integrated conversion
The guided and manual workflows use the same backend as the standalone Converter, with progress and cancellation.
See what is happening
Request metrics, searchable logs, tray operation, and single-instance activation keep the application practical for daily use.
A LITTLE MORE HEADROOM
Know what fits.
Before you load.
Manager estimates model memory with a practical context allowance, so you can choose an appropriate starting point for your available VRAM.
One model occupies the inference GPU at a time. By default, an idle model unloads after three minutes; adjust that to suit your workflow.
Fits
Estimated room for model and context.
Close
Review your context and GPU headroom.
Not recommended
Choose a smaller artifact or reduce memory use.
CONNECT YOUR FAVORITE CLIENT
Your local model.
Your familiar tools.
Use the exact API name shown in Manager to connect an OpenAI-compatible client. Listing models does not load their weights.
Connection guide →CLIENT CONFIGURATION
// Base URL
http://127.0.0.1:8173/v1
// Model
<your-model-api-name>Localhost by default. LAN access is opt-in.
A complete app, with separate model downloads.
The current Manager workspace includes its CPU conversion runtime. Download a GPU engine and model weights separately. Windows releases are currently unsigned.
YOUR NEXT STEP
Put your GPU to work.
Your models, your workspace, your next idea.
Start with NInferEZ Manager.
Windows x64 · Installer or Portable