InvokeAI hosting
A studio-style diffusion workspace — GPU-bound.
InvokeAI is in our catalogue but not yet orderable: it needs GPU capacity, and we would rather list it honestly than sell a CPU plan that disappoints. Tell us what you need and we will tell you where it stands.
InvokeAI, usually just called Invoke, is an Apache-2.0 creative engine for diffusion models developed by the Invoke team and its community. Instead of a page of sliders it offers a unified canvas with layers, in-painting and out-painting, a gallery organised into boards, and a model manager that installs and tracks checkpoints, LoRAs and embeddings. Its node-based workflow editor covers much of the same ground as ComfyUI for people who need repeatable pipelines, while the canvas suits illustrators who want to paint, mask and regenerate in place.
It supports a wide range of model families: Stable Diffusion 1.5, SDXL and 3.5, the FLUX variants, and newer architectures such as CogView 4 and Qwen Image. That breadth is exactly why it wants dedicated hardware. Model files run to several gigabytes each, and generation is GPU work: Invoke’s own requirements start at 4 GB of VRAM for SD 1.5 and rise to 10 GB for FLUX.1. A shared server lets a team keep one model library and one gallery instead of copying weights to every workstation.
What InvokeAI needs
When GPU plans launch: the official ghcr.io/invoke-ai/invokeai container on port 9090, with the INVOKEAI_ROOT directory and Hugging Face cache on persistent storage and one NVIDIA GPU passed through.
What we configure for you
- INVOKEAI_ROOT on persistent storage for models, database and generated images
- HF_HOME cache on persistent storage so downloaded models survive restarts
- GPU passthrough through the NVIDIA container runtime
- Web UI served only through our TLS proxy
- SFTP access to the models folder for uploading your own checkpoints and LoRAs
Which plan fits
No plans are on sale. A useful InvokeAI instance needs at least an 8 GB VRAM GPU for SDXL and 10-12 GB for FLUX models, with 16 GB of system RAM; the largest FLUX.2 Klein model needs 24 GB of VRAM and 32 GB of RAM.
The honest limits
Our current CPU-only plans cannot run local diffusion at a usable speed, so InvokeAI stays in the catalogue as coming soon. API-backed image models work without a GPU, but that is not why most people install Invoke.
What people run InvokeAI for
Illustration and concept art
Paint, mask and regenerate regions on a layered canvas.
Repeatable pipelines
Save node workflows so the same process runs on every batch.
Shared model library
Keep checkpoints, LoRAs and embeddings in one place for a small team.
Licence and version policy
| Item | Position |
|---|---|
| Upstream licence | Apache-2.0 |
| Sold as managed hosting | No |
| Tenancy | Single tenant — one customer per instance |
| Patching | Critical advisories within 24 hours, high severity within 72 |
Our full position on licences, including the applications we decline to sell, is on the licensing page.
Reading on InvokeAI
Works well with InvokeAI
ComfyUI
Node-based image and video generation — waiting on GPU capacity.
on requestDetails →
Stable Diffusion WebUI
The AUTOMATIC1111 interface — GPU-bound, listed for completeness.
on requestDetails →More Speech, image & video we host
InvokeAI hosting: your questions
Can I order InvokeAI hosting now?
What GPU would InvokeAI need?
Where does InvokeAI keep models and images?
Can I use models from Hugging Face or Civitai?
Does InvokeAI have an API?
Tell us what you need
If this is the app you need, say so — demand is how things move up the list.