ONAGER Compute groups supported and planned compute workloads by their intended use. Each entry states the access method and the current boundary of the offering.
“Available in beta” refers to functions for invited tenants; “In preparation” refers to functions that have not been rolled out and have no announced date.
AI templates
Available in beta
Generate images with FLUX.1-schnell through the API of the booked machine. You control the prompt, image size and seed; custom models and custom ComfyUI workflows are not included with this access.
Limits: Beta invitation required; fixed template; no availability commitment.
Available in beta
Send text requests to selected language models through an OpenAI-compatible API. The API address and token apply to the booking period; this access does not run your own code or models on the machine.
Limits: Beta invitation required; fixed model list.
GPU VM
All entries in this group use the prepared Linux VM with the entire graphics card. No confidentiality towards the provider: the provider can technically access the VM.
In preparation
Full GPU access is being prepared as a fresh Linux VM with the entire graphics card, root access, drivers, CUDA and Docker. It is intended for installing and running compatible frameworks, models and containers yourself.
Limits: Not rolled out, no date; licensing and contractual review pending.
In preparation
The prepared Linux VM is intended for adapting compatible models with your own methods and data. Whether a workload fits within one booking depends on the model size, 24 GB of GPU memory and the booking duration.
Limits: Each booking is 1 day (24 hours), extendable up to 7 days; no performance or outcome commitment.
In preparation
Datasets or audio files are intended to be processed with compatible software that you install in the Linux VM. Results are transferred through SSH; ONAGER does not provide a permanently reachable service endpoint.
Limits: No inbound service port, no continuous booking.
In preparation
Compatible GPU code is intended to be developed and tested in a Linux VM with access to the entire graphics card. Access is planned through SSH from a terminal or a supported development environment.
Limits: No display or remote desktop.
In preparation
Compatible rendering jobs are intended to run in the Linux VM without a display. You install the required software yourself and transfer the generated files through SSH.
Limits: Headless only; software and content licences are the tenant's responsibility.
In preparation
The prepared Linux VM is intended for compatible builds, shader tools, headless tests and offline rendering. Tasks that require Windows, DirectX 12, Pixel Streaming, a display or macOS are not part of this offering.
Limits: The tenant checks engine and tool compatibility; no build farm, no inbound services.
Playing games through ONAGER is not offered. Machines are accessed through an API or SSH, without display streaming.
Apple development
In preparation
ONAGER is preparing bookings of an entire Mac M5 for 1 day (24 hours), extendable by 1 day at a time up to 7 days. During the booking, only the tenant uses and controls macOS and the device; the intended purposes are limited to software development, source builds, automated testing, CI and the development tools they require.
Limits: Only within the scope of Apple's macOS licence; no provider jobs during the booking.