Reference
Models
Supported image and video models, their IDs, default GPUs and access requirements.
Catalog
List the live catalog with clockingcloud models ls or GET /v1/models. Use the ID anywhere a model is expected. The default GPU is what gpu: auto selects.
Image models
| ID | Model | Repository | Default GPU | Access |
|---|---|---|---|---|
| z-image-turbo | Z-Image Turbo by Tongyi-MAI | Tongyi-MAI/Z-Image-Turbo | L40S | Open, no token · Apache 2.0 |
| flux-1-schnell | FLUX.1 schnell by Black Forest Labs | black-forest-labs/FLUX.1-schnell | L40S | Open, no token · Apache 2.0 |
Video models
| ID | Model | Repository | Default GPU | Access |
|---|---|---|---|---|
| wan-2.2-5b | Wan 2.2 5B text to video by Wan-AI | Wan-AI/Wan2.2-TI2V-5B | L40S | Open, no token · Apache 2.0 |
Text models
| ID | Model | Repository | Default GPU | Access |
|---|---|---|---|---|
| qwen3-8b | Qwen3 8B by Qwen | Qwen/Qwen3-8B | L40S | Open, no token · Apache 2.0 |
| gpt-oss-20b | gpt-oss 20B by OpenAI | openai/gpt-oss-20b | L40S | Open, no token · Apache 2.0 |
Bring your own model
Bring your own model runs on a Bare GPU only. Start one with clockingcloud gpu up, SSH in with your key, then download and run any Hugging Face model with your own stack. You get the raw GPU, billed per second. ClockingCloud does not install, serve, optimize or tune models you bring, and the managed harness, autoscaling and API apply only to catalog models. If a model you bring is gated, use your own Hugging Face token on the machine. You are responsible for its license.
Licenses
Something unclear or missing? Email hello@clockingcloud.com.