FLUX.2 Flex

Yes. FLUX.2 Flex is an API-delivered model and commercial-use output rights are included in the service terms. There is no separate licensing process required before using generated images in advertising, marketing materials, or published digital content. This is in contrast to the FLUX.2 Dev weights, which carry a non-commercial license for direct self-hosting without a paid commercial license from Black Forest Labs.

Inference steps determine how many iterative refinement passes the model makes during generation. Fewer steps (roughly 6–15) produce faster outputs suitable for concept rounds and rough ideation. The images are usable but may show less fine detail or slightly softer text. Higher steps (30–50) give the model more passes to refine textures, sharpen text, and resolve complex physical detail. The practical advice is to run low steps for concept approval and step up only for the assets going into final production.

Guidance scale controls how strongly the model adheres to your text prompt versus introducing stylistic variation. A higher guidance scale keeps outputs closer to the literal prompt description, useful when you need specific elements placed precisely. A lower guidance scale allows the model more interpretive latitude and can produce more aesthetically interesting results at the cost of some literal accuracy. Most production creative workflows run guidance scale in a mid-range and adjust steps for the primary quality lever.

The core distinction is in the deployment model and customization options. FLUX.2 Dev ships with open weights you can download, self-host, and fine-tune with LoRA. FLUX.2 Flex is API-only, with no weights to manage and no infrastructure to provision. Flex adds the exposed parameter controls (steps, guidance scale) and a stronger typography specialization. Choose Dev if you need to own the model; choose Flex if you want per-request parameter control and managed API access.

FLUX.2 Pro is the fully optimized closed API variant with fixed, tuned settings. Black Forest Labs has already dialed in the parameters for maximum quality and reliability. FLUX.2 Flex exposes those same parameters to you, so you get more control at the cost of needing to understand what you are tuning. For teams that know their creative workload well enough to optimize steps and guidance scale per job type, Flex offers a meaningful advantage. For teams that want the best output with minimal configuration, Pro is simpler.

Yes. It is specifically documented as having stronger typography performance than other FLUX.2 variants. Infographics, promotional banners, UI mockups, and creative with dense on-image copy all perform reliably. At high step counts, fine text, multi-line layouts, and complex typographic structures render with more precision than lower-step runs. For the most text-intensive creative (dense small lettering, complex data tables in infographics), GPT Image 2 remains the benchmark, but FLUX.2 Flex closes the gap significantly for structured, design-led typography.

Up to 4 megapixels (4MP), which maps to approximately 2048×2048 at 1:1 or equivalent pixel counts in other aspect ratios. This is sufficient for most digital applications (social, web, digital OOH, app store assets) at high DPI without additional upscaling. For print workflows requiring very large format output, an upscaling pass may still be needed, though 4MP is a strong starting point.

Yes, up to 10 reference images per generation. This lets you supply product shots, style references, and brand visual assets simultaneously to anchor the output's composition, color palette, and visual language. Multi-reference conditioning is built into the model natively, with no external ControlNet modules or separate pipeline steps required.

FLUX.2 Flex at high step counts produces strong product photography, with accurate surface materials, reflective finishes, fine detail on product features, and clean studio lighting. The combination of physical detail accuracy and 4MP resolution makes it one of the better choices in the FLUX.2 family for e-commerce and brand product imagery. For the absolute ceiling of photorealism on complex materials (water, glass, skin), FLUX.2 Pro or Max may push slightly further; the trade-off is that Flex gives you the step control to tune the result.

FLUX.2 Flex is primarily a text-to-image generation model with multi-reference conditioning. It does not offer the same non-destructive targeted editing workflow as GPT Image 2, where you can change one element while the model holds the rest of the image stable. For editing and iterative refinement workflows on approved assets, GPT Image 2's editing capability is a better fit. For high-quality generation with reference anchoring and parameter control, FLUX.2 Flex excels.

FLUX.2 Flex works well as a two-phase model within a single project: use it at low step counts (6–15 steps) to generate a batch of concept options quickly and get stakeholder alignment, then re-run the approved direction at high step counts (30–50 steps) for production-quality delivery. This eliminates the need to switch between a draft model and a quality model, preserving stylistic consistency between concept and final. In Masonry, you can run both passes in the same canvas without changing models or tools.

FLUX.2 Flex is an AI image generation model from Black Forest Labs, available inside Masonry, the AI creative agent teams use to produce marketing, product, and brand images.

Open a Masonry canvas, pick FLUX.2 Flex from the model selector, and describe the image you need: a product shot, an ad creative, a social post. Masonry generates it, then you refine, edit, and combine FLUX.2 Flex with other models in one workspace.

Yes, you can start generating images with FLUX.2 Flex on Masonry's free tier, then scale up with higher limits and priority processing as your team grows.

FLUX.2 Flex is built by Black Forest Labs. Inside Masonry it runs alongside 50+ image and video models, so your team can pick the right one for each brief without switching tools.

Yes, the prompt gallery on this page shows real images teams have generated with FLUX.2 Flex in Masonry, each paired with the exact prompt you can copy and adapt for your own brand.