Which AI Models Can AI-Flow Connect in One Workflow?

AI-Flow lets you chain models from OpenAI, Google, Anthropic, and Replicate inside a single workflow, so a text model can feed an image model, which can feed a video model, all without leaving the canvas. The page lists text, image, video, music, and speech models together, and the same flow can mix providers — for example, GPT, Claude, and Gemini alongside Flux, Kling, Veo, and Seedance. Because the model list changes over time, treat the in-app list as the source of truth for what's currently available.

Providers AI-Flow pulls from

The site names four model sources explicitly:

• OpenAI — GPT and GPT Image families
• Google — Gemini and Nano Banana families
• Anthropic — Claude (the page references Claude Sonnet 4.6)
• Replicate — Flux, Kling, Veo, Seedance, and others hosted there

The headline claim is "Connect any AI model," and the workflow builder is described as combining models from these providers into end-to-end automations.

Model types you can chain

The listed models span several output types, which is what makes cross-type chaining possible:

A practical chain might be: Gemini or GPT writes a prompt → GPT Image or Flux renders the frame → Kling or Veo animates it → Speech 2.6 adds narration. Each step is a node, and the output of one becomes the input of the next.

How mixing providers works in practice

The page's example prompt — a fashion-film description with tortoiseshell sunglasses and warm studio light — is shown against multiple model names (Nano Banana 2, GPT Image 2, Claude Sonnet 4.6, Seedance 2, Flux 2 Max). That layout reflects the core idea: the same creative brief can be routed to different models, and a single workflow can pass work between them.

Two things to keep in mind when you build a mixed flow:

• Output compatibility. A text model produces text; an image model needs an image prompt or a reference image. Check what each node expects before wiring them together.
• API keys. The page says you can "use your own API keys — or don't," and mentions a free tier with your own keys and no subscription. It does not state which models work without a key, so verify per-model in the app.

Why the model list isn't fixed

The page shows a long, fast-moving roster (GPT Image 2.5 Flare, Seedance 2.5, Kling V2.6, Veo 3.1, and many more). New versions appear and older ones get superseded, so any list copied from a page — including this one — can go stale. Before you commit to a workflow, open AI-Flow and check which models are currently selectable, then confirm the ones your flow depends on are still there.

If your goal is a single pipeline that moves between vendors — say, Claude for scripting, Flux for stills, and Kling for motion — AI-Flow is built for exactly that. If you need a model that isn't in the current list, you'd have to wait for it to be added or route that step through a provider you can call directly.