FLUX Kontext Pro is the fast, iterative tier of Black Forest Labs' FLUX.1 Kontext family — a flow-matching (rectified-flow transformer) model that unifies text-to-image generation and in-context image editing behind one endpoint. Give it a prompt to generate from scratch, or hand it a reference image plus an instruction to edit locally (a single object), globally (a whole scene), match a style, or correct the text rendered inside an image. Because the model reads your input image as context rather than starting from a blank canvas, it preserves faces, objects, and layout across multi-turn edits — the exact point where most editing models drift. It is not a language model: there is no token context window, no function calling, and no document reasoning. Images and text go in; an image comes out.
| Field | FLUX Kontext Pro |
|---|---|
| Maker | Black Forest Labs (BFL) |
| Family / tier | FLUX.1 Kontext — [pro] |
| Architecture | Flow matching / rectified-flow transformer |
| Tasks | Text-to-image · local & global editing · character reference · style reference · in-image text editing |
| Inputs | Text prompt (+ reference image for editing) |
| Output | Image, ~1 MP, multiple aspect ratios |
| Multi-turn | Yes — iterative refinement |
| Open weights | No — [pro] is API-only; only [dev] is open-weight under a non-commercial license |
| Released | May 29, 2025 |
| GPTProto model string | flux-kontext-pro |
| GPTProto endpoint | https://gptproto.com/v1/images/generations |
| GPTProto price | $0.032 / image (official BFL list $0.04) |
On GPTProto this page runs the image-to-image task — a text prompt in, an image out, at $0.032 per image. Text-to-image is FLUX Kontext Pro's signature; on the platform it runs through the editing task rather than this endpoint.
Which tier: Pro or Max?
FLUX Kontext Pro is the value tier: at $0.032/image it's half the price of Max and the fastest of the family, built for text-to-image and everyday edits. Step up to FLUX Kontext Max ($0.064/image) only when in-image typography or exact prompt adherence is the deciding factor — both share the same editing interface and character consistency. See the full tier-by-tier breakdown (and how Max stacks up against gpt-image-2) on the FLUX Kontext Max page.
Switching from Black Forest Labs (drop-in)
Moving from BFL's direct API to the FLUX Kontext Pro API on GPTProto is a small
diff, not a rewrite:
1. base_url → https://gptproto.com/v1/images/generation
2. auth → Authorization: Bearer YOUR_GPTPROTO_KEY (replaces BFL's x-key)
3. model→ keep "flux-kontext-pro" The same GPTProto balance then covers FLUX Kontext Max, Seedream, gpt-image-2 and 200+ other models — one key, one invoice, no separate BFL credit top-ups and no per-provider sign-up.
Prompt tips for FLUX Kontext Pro
Kontext follows edit instructions literally, so specificity beats length:
• Name the subject, don't use pronouns: "the woman with short black hair",
not "her".
• State what must NOT change: "…while keeping the same face and pose".
• To reposition without redrawing: "maintain the original composition".
• For background swaps: "change the background to a beach, keep the person in
the exact same position".
Keep multi-turn sessions short — BFL notes quality can drift after roughly six consecutive edits, so restart from a clean base image if artifacts appear.
