Vercel AI Gateway adds FLUX 3 Image from Black Forest Labs under a single API key

FLUX 3 Image arrives on Vercel AI Gateway

Vercel has added FLUX 3 Image, an image model from Black Forest Labs, to its AI Gateway. According to the Vercel blog, developers can generate images from text prompts and edit existing pictures with the same model, using one API key and without creating a Black Forest Labs account.

The move fits the gateway's core pitch: instead of juggling separate accounts, credentials and billing relationships for every model provider, a single Vercel key unlocks a catalogue of models behind a uniform interface.

What the model supports

FLUX 3 Image covers both creation and editing tasks in one model. For edits, it accepts up to ten reference images, which Vercel says can be used to retouch an existing picture, combine subjects from different sources, or rework a composition altogether.

On the output side, the model offers five resolution tiers, ranging from 768 × 768 up to 4K, with square, portrait and landscape aspect ratios. That spread lets teams trade cost and latency against detail depending on where an image ends up — a small thumbnail rarely needs the same fidelity as a full-width hero graphic.

How to call it from the AI SDK

The integration targets AI SDK 7 or later. Setup amounts to setting the AI_GATEWAY_API_KEY environment variable and using bfl/flux-3-image as the model ID.

For generation, Vercel's example calls the generateImage function with a text prompt describing the desired picture, an aspect ratio such as 1:1, and resolution settings passed through provider options under the blackForestLabs namespace — for instance, requesting a 1K square output.

Editing follows the same call shape. Instead of a plain text prompt, the prompt object carries both a text instruction and an images array holding the input picture, such as the raw bytes of an image returned by a previous generation step. In Vercel's worked example, an earlier render of a vase with wildflowers is re-run with an instruction to change the vase to terracotta while keeping the flowers and the surrounding setting intact.

Terminal and playground access

The same API key and model identifier also work from the command line through AI CLI, Vercel's terminal tool for invoking models. Developers who want to try the model before writing any code can experiment in Vercel's model playground, and the company points to its image generation guide for fuller documentation.

Why it matters

Image generation is becoming a standard feature in web applications, but each provider tends to ship its own SDK, authentication flow and pricing scheme. Routing FLUX 3 Image through AI Gateway collapses that overhead into one key and one request format, with no separate relationship needed with Black Forest Labs.

For teams already building on Vercel and the AI SDK, adding image generation or editing is close to a one-line change, and swapping the model ID is all it takes to move between gateway-hosted models. Because FLUX 3 Image handles both generation and editing in a single model, and spans a wide resolution range, one integration can serve everything from quick drafts to 4K assets rather than forcing different tools for each job.