FLUX.2 fits e-commerce and marketing teams that need photorealistic, brand-accurate product images without a photo shoot, replacing manual studio photography and stock-image licensing. Backed by Black Forest Labs' $3.25 billion valuation, it targets businesses needing consistent, on-brand visuals at scale, not painterly art. It doesn't do video or audio.
FLUX.2 is Black Forest Labs' rectified flow transformer for text-to-image generation, producing native 4-megapixel output across five variants, from an open-weight research checkpoint to a sub-second Klein model for consumer GPUs. Its defining feature is multi-reference conditioning, blending subject, style, and environment from separate source images in one generation call.
Provider: Black Forest Labs · Family: FLUX.2
More about Black Forest Labs on HokAI
Input modalities: text, image · Output: image
About FLUX.2
Black Forest Labs is the German-American AI research company behind FLUX.2, co-founded in 2024 by four Stability AI alumni who built the original Stable Diffusion: Robin Rombach, Andreas Blattmann, Patrick Esser, and Dominik Lorenz. FLUX.2 launched November 25, 2025 as the successor to FLUX.1 (August 2024), the family that became one of Hugging Face's most downloaded image generators. Black Forest Labs raised a $300 million Series B alongside the launch, valuing the company at $3.25 billion. The FLUX.2 family spans five variants built around a shared rectified flow transformer architecture: [max] for cinematic output, [pro] for production APIs, [flex] for developer-tunable inference, [dev] as an open-weight research checkpoint, and [klein] as a compact consumer-hardware option.
FLUX.2 replaces FLUX.1's degraded high-resolution geometry with a redesigned Variational Autoencoder that fixes the texture and anatomy artifacts FLUX.1 showed at its largest output sizes. Black Forest Labs' usage policy bans generating CSAM, non-consensual intimate imagery, and deceptive synthetic media of real people, enforced through automated API content filters and carried into the open-weight license for anyone self-hosting FLUX.2 [dev] or [klein]. The company operates out of Freiburg im Breisgau, Germany, under EU AI Act obligations as a general-purpose AI system provider, and has not published a detailed training-data card specifying dataset composition or a cutoff date.
Pricing
Credit-based pricing: 1 credit = $0.01 USD. Megapixel-based: FLUX.2 [pro] $0.03 first MP + $0.015 per additional MP (full 4MP = $0.075). FLUX.2 [flex] ~$0.01/image. FLUX.2 [dev] ~$0.012/image. FLUX.2 [klein] ~$0.014/image. FLUX.2 [max] priced above Pro; see bfl.ai/pricing for calculator.
Key Features
- Multi-Reference Conditioning: Combines up to 10 reference images (Pro/Max) in a single call, using weighted attention to independently transfer subject identity, style, and environment without cross-contamination between references.
- 4MP Native Resolution: Generates images at up to 4 megapixels natively, doubling FLUX.1's 2MP ceiling. A redesigned VAE ensures clean geometry and texture at high resolution without the degradation seen in first-generation models.
- Exact Hex Color Matching: Accepts specific hex color codes in prompts and reliably reproduces them in outputs, enabling brand-accurate product visuals without manual color correction post-processing.
- JSON-Structured Prompting: Accepts structured JSON alongside natural language prompts, allowing precise specification of camera settings, scene elements, lighting, and composition for consistent commercial image pipelines.
- FLUX.2 [klein] Sub-Second Generation: The 4B and 9B Klein variants generate images in under 1 second on RTX 3090/4070+ (13GB VRAM), enabling real-time image generation for consumer-facing applications.
- Open-Weight Dev Model: FLUX.2 [dev] at 32B parameters is on Hugging Face for non-commercial use, allowing researchers and fine-tuners to experiment with and adapt the full frontier model without API costs.
Pros
- Leads 2026 photorealism comparisons against Midjourney v7, Imagen 4, and GPT Image 2 on product photography and anatomy rendering in multiple independent reviews.
- Multi-reference conditioning (up to 10 refs for Pro/Max) maintains character, style, and environment consistency in a single generation call, a capability absent from most competitors at launch.
- Five-variant family spans consumer-tunable inference to enterprise-grade cinematic output, with open-weight Klein and Dev enabling self-hosted deployment on consumer RTX hardware from 13GB VRAM.
Cons
- Image output only: no audio, video, or text generation, so voice and video workflows require integrating separate model providers.
- No independently published red-team safety report; safety posture is self-attested via the model card, unlike Anthropic and Google, which publish named third-party audit results.
- FLUX.2 [dev] open weights carry a non-commercial license; SaaS deployment above 100,000 images per month requires a paid BFL commercial license, catching teams who prototype with open weights and scale to production.
Benchmarks
A benchmark is an exam, not the job. Scores transfer unevenly between tasks, so weigh the one closest to your workload and read every figure with its source.
Frequently Asked Questions
How much does FLUX.2 cost in 2026?
Pricing runs on Black Forest Labs' megapixel-based credits, where one credit equals a cent. FLUX.2 [pro] is $0.03 for the first megapixel plus $0.015 per additional megapixel, so a full 4MP render costs $0.075. The cheaper tiers run roughly $0.01 per image on [flex], $0.012 on [dev], and $0.014 on [klein], while [max] prices above Pro through BFL's online calculator. Third-party hosts like fal.ai or Replicate sometimes beat BFL's direct rates.
Can you use FLUX.2 without paying?
FLUX.2 does not offer a free API tier: every call through Black Forest Labs is billed per image via credits. Researchers can use the [dev] and [klein] weights on Hugging Face at no cost for non-commercial fine-tuning and experimentation. Scaling that access into a commercial or SaaS product needs a separate paid BFL license once usage passes 100,000 generated images a month.
What are the best alternatives to FLUX.2?
For painterly or cinematic campaign work, Midjourney v7 still leads since FLUX.2 is tuned for photorealism rather than artistic style. GPT Image 2, also on HokAI, is OpenAI's alternative and wins on in-image text rendering thanks to its reasoning-model backbone. Anyone needing video or audio in the same pipeline should look at Black Forest Labs' own FLUX 3, which adds those modalities FLUX.2 doesn't have.
FLUX.2 or Midjourney: which should you pick?
Not universally one or the other: independent 2026 reviews put FLUX.2 ahead of Midjourney v7 on photorealism, product photography, and anatomy rendering. FLUX.2 also offers multi-reference conditioning and exact hex color matching that Midjourney doesn't match at the same fidelity, which matters for brand-consistent commercial work. Midjourney remains the stronger pick for artists chasing a distinctive, non-photographic visual style.
What does it take to start using FLUX.2?
Sign up for an API key at api.bfl.ai, or try any variant with no code first at the BFL playground (playground.bfl.ai). A first image typically comes back through the Python or JavaScript client in well under a minute once you have a prompt and any reference images ready. Developers self-hosting FLUX.2 [dev] need an H100-class GPU for full precision, or an RTX 4090 running FP8 quantization, plus acceptance of the non-commercial license on Hugging Face.
Top Alternatives
- GPT Image 2: Pick GPT Image 2 if you need near-perfect in-image text rendering (OpenAI cites 99% accuracy); pick FLUX.2 for faster, cheaper photorealistic product shots with exact brand colors.
- FLUX 3: Pick FLUX 3 if you need Black Forest Labs' newer video and audio generation in one model; pick FLUX.2 if you only need cheaper, faster still images.