Skip to main content
Agent, media, and Code nodes all run on generative models you install into your Runchat account. Models come from Fal (image, video, 3D, audio) and OpenRouter (text and reasoning), with a small number of additional providers proxied through Fal.

​
Selecting a model

Select a node and click the model button on its toolbar (Swap model) to open Choose a model, filtered to models compatible with that node type. Media nodes show image, video, audio, 3D and other media-producing models. Agent and Code nodes show text and reasoning models. The list on the left of the picker groups models by task (Create image, Edit image, 3D model, Video, Audio and more). Filter by quality (Any, Draft, Balanced, Quality) or Type, and change the Sort to Popular, Name or Cost. Each model lists its credit cost.

​
Installing models

Install new models from the node picker (press Ctrl/Cmd+K or double-click the canvas). There are two ways to add one:
  • Add Fal / Add OpenRouter: Click Add Fal or Add OpenRouter at the bottom of the picker, paste the Model URL and click Install. OpenRouter supports both LLMs and media models; image, video, and audio models install as media nodes automatically.
  • Paste a model URL: Paste a URL from fal.ai/models/... or openrouter.ai/models/... directly into the picker search. Runchat parses the URL and offers to install the model.
OpenRouter image models accept prompts and, where supported, reference images for editing. Video models offer their supported durations, resolutions, and first/last frames. Speech models offer voice selection; other audio models generate audio from a prompt. Generated files are saved to your Gallery. Linked OpenRouter accounts are billed directly; otherwise Runchat credits use the provider’s reported cost. Displayed media prices are estimates and vary with inputs, duration, resolution, and quality. Most models become available immediately after installing. Some models with complex pricing structures need to be manually verified by the Runchat team before they’re enabled, in which case the model appears in the picker but won’t be runnable until verification completes.
Some models use billing methods that aren’t compatible with Runchat credits. If a model you need isn’t supported, get in touch on Discord or use Bring your own keys to bill the provider directly.

​
Picking the right model

Cheaper models are usually faster but less reliable for complex reasoning, multi-step tool calls, or tight quality demands. A common workflow:
  1. Iterate on a cheap, fast model (Flux 2 Klein for images, MiMo for text)
  2. Once the workflow is right, swap to a higher-quality model (NanoBanana Pro, Muse Image, Flux 2 Pro, Claude Opus) for the final pass
See Choosing a model for which image model to start with on each design task, and Credits → Model pricing for a full cost comparison across all default models.

​
Scope and removal

Installed models are scoped per user. They’re available in every workflow you create, regardless of team. Team admins can apply a model whitelist to restrict what team members can install. To remove models, disconnect the corresponding provider integration from your account.

​
Next steps