Black Forest Labs' FLUX 3 Image model can now be called through Vercel's AI Gateway with a single Gateway API key and no separate Black Forest Labs account, according to Vercel's changelog. The same model handles generating an image from text and editing an existing one.
For editing, Vercel says the model accepts up to ten reference images, which can be used to edit, combine subjects or change composition. It lists five resolution tiers from 768 by 768 up to 4K, in square, portrait and landscape shapes, and shows code using the AI SDK version 7 or later, plus a CLI toolA command-line capability an agent can discover and compose through shell commands.Full definition →. This is a distribution and interface note, not a quality evaluation. The model's own release is covered in a separate Intel item.
Key takeaways · AI-distilled
FLUX 3 Image is one model for both text-to-image generation and image editing. On AI Gateway it runs as bfl/flux-3-image with an AI Gateway key, needs no Black Forest Labs account, and requires AI SDK 7 or later.
Editing accepts up to ten reference images, for combining subjects or changing composition. In the AI SDK you pass existing images in PromptInformation or instructions supplied to a model; in a chat, the request you type is often called your prompt.Full definition →.images together with a text InstructionA statement in the context that tells the model how to behave or what to do; module 3's first layer.Full definition →.
Output comes in five resolution tiers from 768 x 768 up to 4K, in square, portrait and landscape aspect ratios; the tier is set through the blackForestLabs provider option, for example resolution: '1k'.
The same key and model ID also work from the TerminalA program that shows text and accepts typed text.Full definition → through AI CLI: install it with npm install -g ai-cli, then run ai image with --model bfl/flux-3-image and an --output directory.
Terms in this piece · Glossary
CLI tool — A command-line capability an agent can discover and compose through shell commands.
Prompt — Information or instructions supplied to a model; in a chat, the request you type is often called your prompt.
Instruction — A statement in the context that tells the model how to behave or what to do; module 3's first layer.
Terminal — A program that shows text and accepts typed text.