Comfy API Launches: Deploy ComfyUI Workflows as Autoscaling Production Endpoints
crystal_alpine · reddit · 2026-10-01
Comfy officially launched Comfy API, tackling the classic pain of moving ComfyUI workflows off your own machine. Feed it a workflow JSON or a Comfy Desktop snapshot, and it resolves custom nodes, models, LoRAs and Python deps (any version can be overridden), pinning everything — including the ComfyUI version — into a Build. You cut an immutable release from the Build and deploy it as an autoscaling endpoint; updating the Build leaves the live release untouched.
Key details:
- Scales to zero when idle, or keep workers warm to skip cold starts
- Per-second GPU billing: RTX PRO 6000 at $4.54/hr, H100 at $6.23/hr, H200 at $7.71/hr, B200 at $11.23/hr
- Requires a paid plan (Standard and up), usage billed separately
- Works with private custom nodes
- The engine stays open source, and Builds remain portable to your own hardware
The team is upfront: if RunPod or Modal works for you, keep using it — this is for when managing the ComfyUI environment itself is the headache.
Related event: ComfyUI Launches Comfy API for One-Click Workflow Deployment(3 posts)→
More from Infra
- AI boom lifts hyperscalers' cost of capital as bond spreads widen ~25bps this year — rohanpaul_ai · 2026-10-01
- DGX Spark price jumps $2k in a week as retailers pull stock — StartupTim · 2026-10-01
- Magnitude (YC S25) Open-Sources Inference Engine with Up to 2x Faster Decode Than llama.cpp — petrusenko_max · 2026-10-01
- Chinese labs reportedly buying expert training data from US data firms like Mercor and SurgeAI — BarnacleBasic1988 · 2026-10-01
- Andrew Chen: The $20K workstation era is back — RTX PRO 6000 and Mac Studio — andrewchen · 2026-10-01
- PyTorch Foundation's Mark Collier: open source is the coordination layer for frontier AI — PyTorch · 2026-10-01