LOCAL
Your machine.
Your engine.
Use your GPU for compatible open models. Your prompts and files stay on your device.
CLOUD CREDITS0
Loading your workspace…
Generate images, video and text on your own hardware. Reach for leading cloud models when you need them. All in one desktop app.
MODEL ECOSYSTEM · NOT A PARTNERSHIP ENDORSEMENT
One workspace, three ways to create.
The execution mode is always your choice.
Use your GPU for compatible open models. Your prompts and files stay on your device.
Access proprietary models through a secure gateway. See the estimated cost before you run.
Get a local recommendation, or approve a cloud step. Never a silent switch to paid compute.
Designed to reduce AI costs by moving
eligible workloads to your own hardware.
Need a proprietary model? Prompt → API → Result. Cloud costs apply only to those steps.
A workspace built to meet you
where your hardware is.
Download the desktop app. Sign in to your account. Your workspace is ready to set up.
> install nodexaThe app checks your GPU, VRAM, RAM and disk space, then recommends suitable models.
> detect hardwareRun locally for zero cloud credits, or choose a premium API model with a visible estimate.
> run your ideaLocal workhorses. Cloud heavyweights.
Know where it runs before you start.
Loading current registry…
VRAM estimates depend on quantization,
resolution and runtime.
Images. Video. Chat. One focused workspace.
From your first prompt to your next pipeline.
Interactive desktop concept. Generation and hardware detection require the desktop application; this website does not run an ML engine.
The desktop app checks GPU, VRAM,
RAM and available storage.
Smaller or optimized image models. Quantization and CPU offload may be needed.
More demanding image workflows, with compatible optimized model variants.
Heavier image and selected video workflows, subject to model and resolution.
This is an estimate. Your browser does not inspect your GPU.
Select a GPU to explore likely local options.
Illustrative, editable inputs. Uses Starter ($8.99/mo). Excludes hardware purchase, taxes and subscription bundles.
Actual savings depend on models, hardware and usage. This is not a provider quote.
Every plan. Every model.
Choose your credits, devices and workflow.
Prices are informational. Purchases, subscriptions, credit packs and plan changes are available only inside the desktop application.
Local generations consume 0 cloud credits. API models consume credits based on current provider cost.
*Local capacity is limited by your hardware. Credit consumption varies by model, resolution, duration and token usage.
Stay on your device.
Processed on your machine.
You decide when an API is used.
Choosing an API model sends the necessary inputs to its provider. Provider data-processing policies apply. Account and anti-fraud data are separate from local creative files.
Know what runs where.
Know what you pay for.
Yes. Local models run on your CPU/GPU and use your RAM and storage. API models run with their provider. The execution mode is shown before every job.
No. Local generation uses zero cloud credits. Your hardware, electricity and storage still have costs.
The initial desktop target is Windows x64 with NVIDIA CUDA GPUs. Exact compatibility must be validated in the desktop runtime. Hardware estimates are guidance, not a support guarantee.
You can choose API models without a powerful local GPU. Smaller local models may work on your CPU, with longer generation times.
The app should recommend a smaller model or an API alternative. Switching to paid cloud processing requires your confirmation.
Local prompts and files remain on your device. If you choose an API model, the necessary inputs are sent to that provider under its data-processing policies.
Eligible workloads use compute you already own. That can reduce cloud spending. Actual savings depend on your hardware, model choices and usage.
Trial access ends after three days. Open the desktop application to manage your plan and continue. Your local files are never deleted. Unused trial credits expire.
That is the hybrid workflow: local steps and explicitly approved cloud steps in one project. Only cloud steps consume credits.
No separate consumer subscription is required for supported API access through this service. Availability depends on our enabled provider integrations.
The Windows installer has not been uploaded yet. Download controls will become active as soon as a verified release is configured. macOS and Linux are planned for later.
Windows 10/11 · 3-day trial · All models · No card required