ComfyUI on ROCm

comfyanonymous / AMD docs. AMD AI-ecosystem ComfyUI page checked 2026-10-12. Instinct-oriented Docker pages exist too — do not treat those tags as R9700 pins.

Node-based image (and video) graphs. AMD documents install on ROCm-enabled PyTorch, then python ComfyUI/main.py → http://127.0.0.1:8188.

What it’s for

Diffusion workflows on the Radeon — not an LLM harness. Does not speak llama-server /v1.

Local endpoint

llama-server base: n/a
Strata base: n/a
Model name: n/a (diffusion checkpoints, not GGUF chat)
API key: n/a

Tool calling

Node graph, not LLM tools.

Long context

n/a

ROCm / Radeon gotchas

Needs a ROCm PyTorch wheel that lists your gfx. This is HIP/ROCm, not llama.cpp Vulkan. Consumer gfx1201 support follows the PyTorch wheel, not this site’s daily-driver GGUF path.

Hybrid local + cloud

Optional cloud nodes exist in the ecosystem; AMD’s page is local ROCm PyTorch.

This is the same question as the calculator’sfully localversushybridtoggle. This page does not quote cloud prices.

Official links

Sources

Copy-paste

More on the connect page. Prove http://127.0.0.1:8080/v1/models first.

ComfyUI on ROCm — not an LLM /v1 client

# AMD AI-ecosystem page: ROCm-enabled PyTorch, then:
python ComfyUI/main.py
# UI: http://127.0.0.1:8188
# Does not speak llama-server http://127.0.0.1:8080/v1 or Strata. Instinct Docker tags are not R9700 pins.

HIP/ROCm PyTorch, not llama.cpp Vulkan. Consumer gfx1201 support follows the PyTorch wheel.

ComfyUI on ROCm — not an LLM /v1 client

# AMD AI-ecosystem page: ROCm-enabled PyTorch, then:
python ComfyUI/main.py
# UI: http://127.0.0.1:8188
# Does not speak llama-server http://127.0.0.1:8083/v1 or Strata. Instinct Docker tags are not R9700 pins.

HIP/ROCm PyTorch, not llama.cpp Vulkan. Consumer gfx1201 support follows the PyTorch wheel.