Flux.1

<p>Flux.1 is the AI image generation model built by Black Forest Labs — the same researchers who originally architected Stable Diffusion, now operating with better resources and a sharper focus. The flagship Flux.1 [pro] delivers 12B parameters of rectified flow transformer magic, producing images with a level of prompt adherence and photorealism that genuinely competes with Midjourney. Three variants cover every use case: Schnell (fast, Apache 2.0 open-source, commercial use), Dev (non-commercial only), and Pro (API-only, paid). Free to start with Schnell; Pro API runs ~$0.04/image.</p>

<h3>The Black Forest Labs Origin Story</h3>
<p>Black Forest Labs was founded in 2024 by three former Stability AI researchers — Robin Rombach, Andreas Blattmann, and Patrick Esser — the same people who built Stable Diffusion at LMU Munich under Björn Ommer. They left Stability, raised $31 million from Andreessen Horowitz, and came back with Flux.1, a model designed from the ground up with everything they wished they’d had the first time.</p>
<p>That institutional knowledge shows. Flux isn’t trying to replicate Stable Diffusion with better weights — it’s a completely different architecture (rectified flow transformer, 12B parameters vs SDXL’s ~3.5B) that happens to share some philosophical DNA.</p>
<h3>The Three Variants: Read This Before You Start</h3>
<p>Flux.1 is not one model. Understanding the licensing is non-negotiable before you build anything with it.</p>
<p><strong>Flux.1 [schnell]</strong> — German for “fast.” Distilled to 4 steps (most models need 20-50), generating images in under 2 seconds on modern hardware. Apache 2.0 licensed — free for commercial use, including API products you’re building. This is where most developers should start.</p>
<p><strong>Flux.1 [dev]</strong> — Quality between Schnell and Pro via guidance distillation. The catch: non-commercial use only. I’ve seen teams build entire pipelines on Dev before someone noticed the license. Don’t be that team. If you’re commercial, use Schnell (free) or Pro (paid).</p>
<p><strong>Flux.1 [pro] / 1.1 Pro</strong> — Flagship, API-only. No public weights released. Accessed via Black Forest Labs directly or providers like Replicate and fal.ai at ~$0.04-0.06/image. 1.1 Pro is the current production standard with improved quality and speed. This is the tier that competes with Midjourney directly.</p>
<h3>What Actually Impressed Me: The Text Rendering</h3>
<p>Getting legible, correct text inside generated images has been a persistent failure mode for image models for years. DALL-E 3 improved it. Midjourney v6 improved it. Flux.1 Pro essentially solved it — short text strings render correctly and readably in a way that still fumbles on most competitors. If your use case involves product packaging mockups, UI wireframes, signs, labels, or anything with typographic elements, this is a genuine differentiator.</p>
<p>On photorealism, Flux.1 Pro sits at the level I’d previously associated only with Midjourney’s top tier. Skin texture, lighting consistency, environmental detail — all hit the bar where I’d hand the output to a client without hesitation.</p>
<p>The caveat: Midjourney still leads on certain artistic and painterly styles. Years of community refinement are baked into Midjourney’s defaults. For anything that needs to look real, Flux has closed the gap. For pure creative or stylized work, Midjourney retains the edge.</p>
<h3>Flux vs the Field</h3>
<p><strong>vs Midjourney:</strong> Midjourney leads on artistic/aesthetic style and ease of use — beautiful images with minimal effort via Discord. Flux wins on API access, text rendering accuracy, and cost (Schnell is free commercially). Choose Flux for product pipelines; choose Midjourney for pure creative exploration.</p>
<p><strong>vs Stable Diffusion XL:</strong> SDXL’s advantage is ecosystem maturity — more LoRAs, more community knowledge, more tutorials accumulated over years. Flux wins on raw quality and prompt adherence. SDXL is still the better choice if you need deep customization via LoRAs or ComfyUI workflows; Flux is the better choice if quality and speed are the priority.</p>
<p><strong>vs DALL-E 3:</strong> DALL-E 3 integrates tightly with ChatGPT Plus ($20/month) for convenience. Flux Pro wins on raw image quality and text rendering. DALL-E wins on ecosystem simplicity — no API keys to manage, no pricing to optimize.</p>
<h3>Who Should Use Flux.1</h3>
<p><strong>Developers</strong> building AI image generation into products need the API access and commercial licensing (Schnell) that open-source provides. <strong>Teams needing text-in-image</strong> — product mockups, UI designs, marketing assets with labels — will find Flux’s text rendering genuinely solved in a way competitors still struggle with. <strong>Researchers and evaluators</strong> can start with Dev (free, non-commercial) to benchmark against their existing pipelines. <strong>High-volume content teams</strong> should start with Schnell API (~$0.003/image) and only upgrade to Pro when quality requirements demand it.</p>
<p>Less ideal for: casual users who want polished image generation with zero setup (use Midjourney via Discord); anyone who needs the deepest LoRA customization ecosystem (use SDXL with ComfyUI).</p>