FLUX.2
by Black Forest Labs
Production-grade images up to 4MP, with up to ten reference images.
What it is
FLUX.2 is Black Forest Labs' image model, officially "a production-grade AI image generation and editing model with 4MP photorealistic output and multi-reference control". The word production is doing real work in that sentence: this is aimed at output you ship, not output you screenshot.
The feature that sets it apart is multi-reference. It takes up to ten reference images at once, which is what you need when a shot has to match an existing product, an existing person and an existing set of brand colours all at the same time. Black Forest Labs also calls out reliable spatial reasoning: accurate positioning, realistic physics and coherent lighting.
It renders any aspect ratio, works from drafts as small as 400 pixels square, and handles complex typography and interface mock-ups. In this studio it runs on both the image and the edit side, and it is one of the models that takes your own LoRA.
What it is good at
Multi-reference control with up to ten images simultaneously. The realistic case: this product, this person, this palette, in one shot.
Black Forest Labs calls it "production-ready text" , complex typography and UI mock-ups that actually work. Rare, and immediately obvious when it is missing.
Reliable spatial reasoning: accurate positioning, realistic physics and coherent lighting. Fewer objects floating a centimetre above the table.
Train a look or a likeness once and apply it here. That is how a brand style stops being something you re-describe every single time.
Where it struggles
- Black Forest Labs publishes no list of limitations, so there is none here. What we will say is what we know: this is a family with several tiers, and the cheapest is not the one the marketing describes.
- The variants differ in more than speed. Max is the strongest at following a prompt, Pro trades a little of that for speed, Flex is the precise one. Picking on price alone is how people end up disappointed by a model that was never the one being praised.
How to prompt it
Write it like a brief. It handles long input , 32,000 tokens of it , so the failure mode here is being too vague, not too long. Name the subject, the framing, the light and the finish, and attach references for anything that has to match something real rather than describing it in words.
An example that shows what that means
A ceramic pour-over cone on a pale oak counter, shot straight on at eye level with a 50mm lens at f/4, morning light from a window to the right and a soft bounce from the left, matte glaze catching one clean highlight, neutral photographic colour, a shallow shadow anchoring it to the surface.
What we offer here
Straight from the catalogue, so this list never goes stale. The price is the provider’s published rate with our markup on top, and it is on the button before you run anything.
Questions
How many reference images can I use?
Up to ten at once, per Black Forest Labs. That is the feature to reach for when a shot has to match several real things at the same time.
Which variant should I pick?
Max follows a prompt most closely, Pro is the fast one at high quality, Flex is built for precision and control. The Studio shows what each costs before you run, so the trade is visible.
Can I use my own trained style?
Yes. This is one of the families in the Studio that accepts a LoRA, so a look you trained once applies to everything you make with it.
Is it good for text in images?
It is one of the reasons to choose it. Black Forest Labs specifically calls out complex typography and UI mock-ups.