Running ComfyUI on ComfyOnline: Understanding the Three Operational Modes
ComfyOnline offers users flexible options for executing their ComfyUI workflows, catering to various needs and use cases. The platform currently provides three distinct operational modes:
1. Shared Serverless Mode
Billing: Charged only during the generation (inference) process.
Resource Usage: Your ComfyUI workflows are executed on GPUs allocated from a shared pool of machines managed by the platform.
Hardware: The shared pool primarily consists of NVIDIA RTX 4090 GPUs (24GB VRAM), with a limited number of L40S GPUs (48GB VRAM) also available.
Default: This is currently the default operational mode on ComfyOnline.
2. Independent Deployment Serverless Mode
Billing: Charged only during the generation (inference) process.
Resource Usage: This mode provides a dedicated execution environment for specific, user-defined ComfyUI workflows. Your chosen workflow is independently deployed and routed to a set of dedicated machines. The platform currently defaults to configuring 2-3 independent instances for these deployments to ensure stability and performance.
Hardware: Supports a wider range of high-performance GPUs, including NVIDIA RTX 4090 (24GB), L40S (48GB), A100 (80GB), and H100 (80GB).
Use Case: Offers enhanced stability and predictable performance, making it highly suitable for API integrations, production workloads, and users requiring guaranteed resources for specific workflows.
Requirement: Requires a subscription to the Startup Plan or Business Plan to enable.
3. Dedicated Machine Mode (Continuous Billing)
Billing: Billing starts the moment the machine is launched and continues until the user manually shuts it down. This is a continuous billing model.
Resource & Control: You are allocated a fully dedicated and controllable virtual machine.
Flexibility: Provides maximum flexibility, allowing you to freely install custom ComfyUI nodes, models, or any other software and configurations required for your specific needs.
Description: This mode is designed for users who require complete control over their ComfyUI environment. If you need to install numerous custom extensions, specific models, or have unique setup requirements, a dedicated machine offers the freedom to customize your workspace fully. Billing is continuous as the machine resources are reserved for you exclusively while it is running.