FLUX.2 [klein] 4B

FLUX.2 [klein] 4B is a 4 billion parameter rectified flow transformer designed to bring text-to-image generation and editing together in a single compact architecture. The model targets real-time and interactive use cases, with end-to-end inference described as low as under a second, and it is positioned as the fastest distilled member of the FLUX.2 [klein] family. Its small parameter count and low VRAM footprint, reportedly as little as 13 GB, allow it to run on consumer hardware such as the RTX 3090 and 4070, making it well suited to local development, edge deployment, and production pipelines that need a balance of speed and quality.

The model is released under the Apache 2.0 license, reflecting an open approach to weights and redistribution. Beyond straightforward text-to-image prompts, FLUX.2 [klein] 4B supports multi-reference editing, so users can guide outputs with several reference images rather than only text. This combination of unified generation and editing in a 4B parameter footprint makes the model a practical choice for applications that require quick, iterative visual content without committing to a much larger generative model.