Newsletter image

Subscribe to the Newsletter

Join 10k+ people to get notified about new posts, news and tips.

Do not worry we don't spam!

By pressing the Subscribe button, you confirm that you have read and are agreeing to our Privacy Policy and Terms of Use

Search

GDPR Compliance

We use cookies to ensure you get the best experience on our website. By continuing to use our site, you accept our use of cookies, Privacy Policy, and Terms of Service.

Black Forest Labs - Image Generation

Black Forest Labs Drops FLUX.2

Black Forest Labs has launched FLUX.2, an advanced image generation model offering native 4MP output, multi-reference editing, improved prompt adherence, and high-quality typography. It supports any aspect ratio and generates images in under 10 seconds. Key features include superior text rendering, anatomy, and scene composition, along with color matching and efficient VRAM usage. The FLUX.2 lineup includes several models, each tailored for different needs, with varying levels of access and licensing. The minimum system requirements are specified for optimal performance. FLUX.2 is recognized as a leading open-weight image model, suitable for various creative and commercial applications. The model and resources are available on platforms like Hugging Face and GitHub.
2025-11-28
Updated 2025-11-28 11:45:37

Black Forest Labs just unleashed FLUX.2 — and it’s a monster. Released November 25, 2025, the new family pushes open-weight image generation to frontier level with native 4MP output, multi-reference editing (up to 10 images for consistent characters and styles), dramatically better prompt adherence, and production-ready typography that finally doesn’t look broken. It supports any aspect ratio, reliable low-res drafts from 400px², and sub-10-second generation speeds. Check out the official announcement and model details for all the specs.

Why Everyone Is Talking About FLUX.2

  • Native 4-megapixel resolution (no upscaling hacks; any aspect ratio)
  • Multi-reference control — combine up to 10 images for identity consistency across assets
  • Best-in-class text rendering and font handling for infographics and UI mockups
  • Superior anatomy, lighting, complex scene composition, and spatial reasoning (physics, perspective)
  • Generative editing: replace elements (e.g., objects, characters), adjust poses/proportions, expand/shrink scenes
  • Exact color matching via hex codes; grounded in real-world knowledge
  • 40% less VRAM + 40% faster inference via NVIDIA FP8 optimization
  • Runs locally on consumer GPUs with ComfyUI templates

The Full FLUX.2 Lineup

Model Size Strength License / Access
FLUX.2 [pro] ~32B State-of-the-art quality & maximum speed Proprietary API only
FLUX.2 [flex] ~32B Maximum precision, readable text, full creative control API only
FLUX.2 [dev] 32B Frontier performance for deployment; full weights + editing Hugging Face (non-commercial research)
FLUX.2 [klein] Distilled Edge & consumer friendly Apache 2.0 (coming soon)
FLUX.2 VAE Improved latent space Apache 2.0 (available now)

Minimum System Requirements (FLUX.2 [dev])

Resolution GPU VRAM (FP8) GPU VRAM (BF16/FP16) Recommended GPU
1024×1024 12 GB 20 GB RTX 3060 12 GB / RTX 4070+
2048×2048 (2MP) 16 GB 28 GB RTX 4080 / 4090 / A6000
4MP (e.g. 2688×1536) 20–24 GB 38+ GB RTX 4090 / A100 40 GB+

Requirements measured with ComfyUI + official FP8 weights. Using the new FLUX.2 VAE and torch.compile reduces memory further by ~10–15%. Supports 32K text input tokens.

Independent benchmarks already crown FLUX.2 [dev] as the strongest open-weight image model ever released — outperforming SD3, AuraFlow, and every previous FLUX variant by a wide margin in both generation and editing tasks, closing the gap with real photography in details like fabric textures and architecture.

For creators, designers, and researchers, FLUX.2 isn’t just another model drop. It’s the moment open-source visual AI finally caught — and in many areas surpassed — the closed giants. Ideal for marketing (consistent campaigns), product viz (photoreal renders), UI/UX (text mockups), and e-commerce (scaled photography).

Grab the weights on Hugging Face, fire up ComfyUI, and start fluxing. Official inference code and examples: GitHub repo. Try it in the playground.

Prev Article
OpenAI’s gpt-oss Models: A New Era for Open-Source AI
Next Article
OpenThinker-32B

Related to this topic: