Flux.2 Flex
What is Flux.2 Flex?
Flux.2 Flex is the adjustable image model in Black Forest Labs' Flux.2 family. It exposes steps, guidance, seed, and prompt upsampling alongside the shared prompt, size, aspect-ratio, and reference-image inputs. That extra control is useful when a team needs to test settings instead of accepting a fixed generation setup.
Output can be set to 0.6, 1, 2, or 4 megapixels. Seven aspect ratios cover ultrawide, landscape, square, and portrait work. A request can also carry up to ten reference image URLs, which makes the model suitable for work that starts from product shots, visual references, or an approved design direction.
Key features of Flux.2 Flex
Manual generation controls
Set steps and guidance for each request. This makes controlled setting tests possible before a team settles on a repeatable setup.
Seed and prompt upsampling
Fix a seed for closer comparisons between runs, or leave it random for more variation. Prompt upsampling can expand a short prompt before generation.
Four output sizes
Choose 0.6, 1, 2, or 4 megapixels. Draft at a smaller size, then use a larger result once the prompt and settings are settled.
Seven aspect ratios
The available shapes run from 21:9 landscape through 1:1 square to 9:21 portrait, covering common site, social, presentation, and display layouts.
Up to ten references
Add reference image URLs to guide the subject, layout, or visual direction without moving to a separate model page.
Best for
Look development
Compare steps and guidance values while keeping the prompt and seed stable.
Concept art
Test several visual directions before choosing a final setup.
Reference-led images
Use product shots, layouts, or mood boards as input references.
Repeatable tests
Fix the seed when comparing one setting or prompt change at a time.
Variants
Flux.2 Flex shares its core image inputs with Pro and Max. The main difference is the amount of control exposed for each request.
Flex
Use Flex when steps, guidance, seed, or prompt upsampling are part of the test. It is the family member for setting sweeps and repeatable comparisons.
Pro
Use Pro for standard image jobs that do not need manual diffusion settings. It keeps the shared aspect-ratio, output-size, and reference-image choices.
Max
Use Max for final image work where the model's highest output setting matters more than manual control.
Use cases
For concept art, keep the prompt and seed fixed while changing steps or guidance, then compare the results against the same brief. For a product campaign, attach approved product shots and generate the required landscape, square, and portrait crops from one visual direction. A design team can start at 0.6 or 1 megapixel while testing ideas, then move the chosen setup to 2 or 4 megapixels. The same controls also fit repeatable QA: change one prompt phrase or one setting per run, record the result, and keep the request that best matches the target.
FAQs
People also ask
Flux.2 Flex is the adjustable model in the Flux.2 family. It adds steps, guidance, seed, and prompt upsampling controls to the shared image-generation inputs.
The API offers 0.6, 1, 2, and 4 megapixel output across seven aspect ratios, from 21:9 landscape to 9:21 portrait.
Yes. A request can include up to ten reference image URLs to guide the subject, layout, or visual direction.
Use a fixed seed when you need to compare prompt or setting changes against a more consistent starting point. A random seed is better when you want wider variation.
Use Pro when you do not need manual steps or guidance controls. It keeps the same core size, ratio, and reference-image options with fewer settings to manage.
