Templates

One-click GPU templates: ComfyUI, Ollama, vLLM, Stable Diffusion WebUI, PyTorch and more. Pick a template, pick a machine, connect.

• PyTorch + JupyterLab (notebook, official) - Ubuntu 22.04, CUDA 12.4, PyTorch (cu124), JupyterLab and SSH. The default interactive image.
• TensorFlow + JupyterLab (notebook, official) - Ubuntu 22.04, CUDA 12.4, TensorFlow (and-cuda), JupyterLab and SSH.
• Plain CUDA + SSH (general, official) - Bare CUDA 12.4 container with SSH only: bring your own stack.
• ComfyUI (image gen, official) - Node-based Stable Diffusion / Flux UI on port 8188, plus SSH. · needs 8 GB VRAM
• Stable Diffusion WebUI (A1111) (image gen, official) - AUTOMATIC1111 Stable Diffusion WebUI on port 7860, plus SSH. · needs 6 GB VRAM
• Ollama (llm, official) - Ollama server on port 11434 (OLLAMA_HOST=0.0.0.0), plus SSH. Pull any model with `ollama pull`. · needs 8 GB VRAM
• vLLM (OpenAI-compatible) (llm, official) - vLLM OpenAI-compatible server on port 8000; set MODEL to a Hugging Face model id. · needs 16 GB VRAM
• Axolotl fine-tune (job) (finetune, official) - Runs an Axolotl training config as a job and exits when done. · needs 24 GB VRAM