HeadsUpAI

ComfyUI Adds OpenRouter for Unified Access to Frontier Creative Models

ComfyUI, a node-based orchestration platform for image and video generation, released an official OpenRouter partner node. This integration provides a unified gateway to frontier models like Claude and Gemini. The interface reconfigures dynamically: vision models expose up to 20 image slots, while reasoning models display effort controls.
Vision capacity
Up to 20 images per request
Reasoning effort
Off, Low, Medium, and High
Web grounding
Perplexity Sonar models

The update expands on ComfyUI's Gemma 4 integration by further broadening the platform's multimodal reasoning capabilities. By centralizing access, creators can leverage specific provider strengths—such as Perplexity for web-grounded context—within a single, cohesive workflow.

You can deploy the node in local or cloud ComfyUI environments using official templates. The node includes a real-time pricing badge that implements the transparency of OpenRouter's model comparison tool. Multimodal support currently includes images and video, with specific slots for models like Qwen 3.6 and Gemini 3.5 Flash.

View the full update on cloud.comfy.org
ComfyUI
ComfyUI
@ComfyUI
X

ComfyUI just added @OpenRouter support. Instead of being locked into a single LLM, you can now access 20+ models directly inside Comfy. More flexibility, less friction, same workflow. Links to the workflows below👇 https://t.co/Y2fonfWApT

22retweets162likes
View on X

Still wondering? A few quick answers below.

The OpenRouter LLM node is an official partner integration for ComfyUI that connects the visual orchestration platform to over 20 different large language models. It allows creators to use frontier intelligence from providers like Anthropic, Google, and OpenAI to analyze images, generate prompts, or provide web-grounded context directly within their creative workflows.

You can access a curated list of over 20 models including Claude, GPT-4, Gemini, and Grok. The integration also supports open-weight models like DeepSeek, Qwen, and Mistral, as well as specialized models like Perplexity Sonar for web search. The node interface reconfigures itself automatically to show the specific inputs and capabilities available for whichever model you select.

Yes, the node supports multimodal inputs for models that have vision capabilities. You can use up to 20 image slots for models like Claude or GPT-4o to analyze reference visuals. Additionally, specific models such as Gemini 3.5 Flash and Qwen 3.6 Plus support video inputs, allowing for advanced video analysis and prompt engineering within ComfyUI.

The OpenRouter LLM node is available as an official partner node for both local and cloud versions of ComfyUI. You can get started by using the official workflow templates provided by the Comfy-Org team on GitHub or through the Comfy Cloud interface. Users will need an OpenRouter API key to authenticate and connect to the various model providers.

Pricing is determined by the specific model you choose to run through the OpenRouter API. To help manage costs, the ComfyUI node features a dynamic badge that displays the real-time token pricing for the currently selected model. This provides transparency on input and output costs directly within the workflow as you switch between different frontier or open-weight models.

Share this update