Skip to main content
The ComfyUI SGLDiffusion plugin has two modes. Server mode talks to a standalone sglang serve process over HTTP. ComfyUI sends prompts and receives images or video. This is the same path as the OpenAI-compatible API. Integrated mode keeps ComfyUI’s CLIP, VAE, and sampler loop. SGLang loads only the DiT and runs one forward per sampler step. The worker starts the native Flux / Qwen-Image / Z-Image / MiniMax-H3 pipeline under --comfyui-mode. A single-file ComfyUI .safetensors (or a GGUF overlay) is loaded through a checkpoint spec. Each apply_model call is translated by a per-model adapter. Integrated mode does not ship a separate comfyui_* pipeline class per model.

​
Install the plugin

  1. Install sglang[diffusion]. See Installation.
  2. Copy python/sglang/multimodal_gen/apps/ComfyUI_SGLDiffusion into ComfyUI’s custom_nodes/ directory.
  3. Restart ComfyUI.
Example workflows live next to the plugin under workflows/.

​
Integrated mode

  1. Load the DiT with SGLDiffusion UNET Loader.
  2. Set num_gpus, tp_size, model_type, or compile flags with SGLDiffusion Options.
  3. Connect the loaded model to a standard ComfyUI sampler.
Supported integrated-mode families: Flux, Qwen-Image, Z-Image, MiniMax-H3. Qwen-Image edit is experimental.

​
How a sampler step reaches the worker

The plugin README has the architecture diagram. The hop is:
  1. A model adapter packs ComfyUI tensors into an SGLang Req.
  2. Local ZMQ replaces CUDA tensors with IPC handles so latents stay on GPU.
  3. Rank 0 materializes the handles. Multi-rank --comfyui-mode then detaches CUDA tensors and broadcasts them over NCCL. The general SP / CFG / TP path is still the original broadcast_pyobj.
  4. The worker pipeline keeps transformer plus a pass-through scheduler. After the first step, conditioning stays in a worker session; later steps send latents and the timestep.
Integrated mode currently supports: MiniMax-H3 example graphs: minimax_h3_t2v_sgld.json (T2V / I2VA / FL2VA on the FL2VA DiT), minimax_h3_r2v_sgld.json (reference-to-video; use a ref2va file), minimax_h3_t2v_sgld_upscaler.json (two-pass 3D ×2). Set SGLDOptions.model_type to minimax_h3. On 24 GB, keep the 32B CLIP on CPU and prefer pruned BF16, GGUF, or kitchen_int8 with DiT layerwise offload. See the MiniMax-H3 cookbook. To add a model, register a checkpoint spec under runtime/loader/comfyui_checkpoints/ and a ComfyUIModelAdapter. Do not add another ComfyUI pipeline class.