Black Forest Labs has just introduced the FLUX.2 [klein] model family, targeting developers, creators, and research teams that demand fast, high-quality image generation and editing. The launch includes both the 4B and 9B model variants, with the 4B model released under Apache 2.0 for maximum accessibility and the 9B under the FLUX Non-Commercial License. These models are public and available now for use and testing, with demos open to all.
Introducing FLUX.2 [klein]. Blazing fast. Beautiful.
— Black Forest Labs (@bfl_ml) January 15, 2026
Generate stunning images in under a second while maintaining exceptional quality.
Great for fast editing, changing styles, and developing ideas from 0 → 1.
Available via API, or run it locally - Klein 4B under Apache 2.0,… pic.twitter.com/6MlLm6OSXS
FLUX.2 [klein] stands out for its sub-second inference speeds, capable of generating or editing images in less than 0.5 seconds, while fitting on consumer GPUs with as little as 13GB VRAM. Technical highlights include:
- Unified architecture supporting text-to-image, image editing, and multi-reference generation.
- Quantized FP8 and NVFP4 versions that reduce VRAM usage by up to 55%.
Unlike previous releases, the [klein] models deliver photorealistic outputs with high diversity and production-ready APIs. Benchmarks indicate the 9B model matches or surpasses much larger competitors in both speed and visual quality, and the quantized models offer even broader hardware support.
Black Forest Labs is known for pioneering efficient and accessible AI models, and with FLUX.2 [klein], the company is advancing its mission of powering real-time, agent-ready visual intelligence. Early testers have praised the combination of speed, versatility, and open access, noting the potential for rapid prototyping and integration into diverse workflows.
