FLUX — The Sharpest AI Image Model on the Planet
12-billion parameter Rectified Flow Transformer. Industry-leading prompt adherence, native text rendering, and an open-source core. From the researchers behind Stable Diffusion.
What is FLUX?
FLUX is a family of <span class="hl-pill">state-of-the-art image generation models</span> developed by <span class="hl-pill">Black Forest Labs</span>, founded by the core team behind Stable Diffusion. Built on a novel <span class="hl-pill">Rectified Flow Transformer</span> architecture with 12 billion parameters, FLUX sets a new standard for how closely an AI model can follow a text prompt — turning complex, detailed descriptions into precisely matching visuals with a fidelity no diffusion model has matched before.
Where other generators approximate your intent, FLUX executes it. Its <span class="hl-pill">prompt adherence</span> engine processes every clause of your description, preserving spatial relationships, object counts, lighting conditions, and stylistic nuance. FLUX is also the breakthrough in <span class="hl-pill">AI typography</span>: for the first time, an image model can render legible, stylistically appropriate text inside generated images — opening up posters, product labels, infographics, and branded content to AI workflows.
The FLUX family spans six specialized variants — from the <span class="hl-pill">Apache 2.0 open-source</span> Schnell model (runs locally, commercially free) to the 4-megapixel Ultra and the groundbreaking Kontext model for <span class="hl-pill">in-context image editing</span>. Every variant shares the same transformer backbone; each is tuned for a different balance of speed, quality, and capability. Use them all from one platform.
What Makes FLUX Different
Prompt Adherence
FLUX doesn’t paraphrase your prompt — it executes it. Object counts, spatial positions, lighting descriptions, style references: the Rectified Flow Transformer processes every token of your instruction. Benchmark after benchmark places FLUX at the top for instruction-following fidelity. If you typed it, FLUX renders it.
Typography & Text-in-Image
Legible text inside AI-generated images has been an unsolved problem since diffusion models launched. FLUX solves it. Posters with readable headlines, product packaging with correct copy, infographics with accurate labels — FLUX renders text that belongs in the image, not just pixels that vaguely resemble letters.
Open Source (Apache 2.0)
FLUX.1 Schnell is fully open-source under the Apache 2.0 license — use it commercially, run it locally, fine-tune it, integrate it into any product. No royalties, no restrictions, no lock-in. Black Forest Labs also open-sourced the model weights so the research community can build on top of the most capable image model ever released.
7 Models, One Family
No other provider offers this range. Schnell for instant drafts in 2 seconds, Dev for research and fine-tuning, Pro for commercial production, 1.1 Pro for 6× faster output at the same quality, Ultra for 4-megapixel print-ready images, and Kontext for context-aware editing. Pick the model that fits your task — speed, quality, or precision.
Key Features
Rectified Flow Transformer
A fundamentally different architecture from diffusion models. Rectified Flow learns straighter generation paths, delivering sharper details and more coherent compositions with fewer sampling steps — speed and quality in one.
Industry-Best Prompt Fidelity
Top-ranked on ELO-based human preference benchmarks. FLUX processes complex, multi-clause prompts with spatial reasoning and attribute binding that other models lose track of. What you write is what you get.
Native Text Rendering
Render readable headlines, labels, captions, and typographic compositions directly in generated images. Logos, posters, product mockups, and infographic content — FLUX handles text as a first-class visual element.
Open Source (Apache 2.0)
FLUX.1 Schnell is freely available for commercial use. Download the weights, run locally, fine-tune for your domain, build products without licensing fees. The most capable open-source image model ever released.
4MP Ultra Resolution
FLUX.1 Ultra generates images at true 4-megapixel resolution — up to 2048×2048 pixels. Print-ready quality, fine-detail product photography, and large-format creative work without external upscaling.
Kontext In-Context Editing
FLUX Kontext understands the existing content of an image and edits it selectively. Change a background, swap an outfit, alter lighting — without touching what you want to keep. True image-to-image with semantic understanding.
How It Works
From prompt to pixel in three steps
Write Your Prompt
Describe your image in plain language. FLUX handles complex, multi-clause descriptions with spatial reasoning. The more detail you provide, the more precisely FLUX matches your intent.
FLUX Generates
The Rectified Flow Transformer processes your prompt through 12 billion parameters, building your image along learned straight-path trajectories — faster and sharper than traditional diffusion sampling.
Refine or Edit
Not quite right? Adjust your prompt, switch to a higher-quality variant, or use FLUX Kontext to edit specific elements of the generated image while preserving everything else.
The FLUX Model Family
Six specialized variants — one for every use case and budget
Benchmark Performance
FLUX vs the competition — independent evaluation results
Who Uses FLUX — and How
Product Photography
Generate studio-quality product shots from text descriptions. Marble surfaces, dramatic lighting, lifestyle contexts, seasonal campaigns — FLUX’s prompt fidelity ensures the product is always the hero of the frame.
Brand & Marketing Assets
Campaign visuals, social media graphics, email headers, and ad creatives — all with accurate brand colors, readable text overlays, and consistent style. FLUX handles the full spectrum of marketing production needs.
Infographics & Data Viz
Create information-dense visual content with readable labels, accurate statistics, and clean layout. FLUX’s text rendering makes it the first AI model that can generate infographics you actually want to publish.
Fashion & Apparel Design
Visualize clothing designs, textures, and colorways on realistic figures. Rapid concept iteration for collections, lookbook photography, and e-commerce imagery — all before sampling a single physical garment.
Game & Concept Art
Character sheets, environment concepts, prop design, UI elements — FLUX handles the full visual development pipeline for games and interactive media. Fine-tune FLUX Dev on your own art style for a truly custom asset pipeline.
Architecture & Interiors
Photorealistic architectural visualizations, interior mood boards, and renovation concepts. FLUX’s spatial reasoning handles complex room layouts, lighting angles, and material textures with precision that surprises even architects.
Self-Hosted & API Workflows
FLUX.1 Schnell’s Apache 2.0 license means you can embed it directly into your product, run it on your own infrastructure, and generate at scale with zero per-image licensing costs. The only top-tier model you can truly own.
FLUX vs Other Image Models
See how FLUX stacks up against the top alternatives — all available on our platform.
| Features |  FLUX Black Forest Labs |  Midjourney V7 Midjourney |  GPT Image OpenAI |  NB Pro Yesmind |
|---|---|---|---|---|
| Architecture | Rectified Flow Transformer | Diffusion | Autoregressive (GPT-4o) | Gemini 3 Pro |
| Parameters | 12B | Undisclosed | Undisclosed | Undisclosed |
| Prompt Adherence | Best-in-class (ELO #1) | Good | Excellent | Excellent |
| Text Rendering | Excellent (92%) | Poor (~30%) | Excellent | Best (94%) |
| Open Source | ||||
| Max Resolution | 4MP (2048×2048) | 1024×1024 | 4096×4096 | 4096×4096 |
| Image Editing | FLUX Kontext | Vary / Remix | Conversational | |
| Fine-Tuning | Limited (--sref) | |||
| Commercial License | Apache 2.0 (Schnell) | Paid plans only | Paid plans only | Paid plans only |
| Best For | Prompt fidelity, typography, open-source | Artistic aesthetic | Conversational editing | Studio-quality output |
Prompt → Result Gallery
Generated with FLUX Pro 1.1 — professional AI image generation.
