FLUX.2 [pro]

FLUX.2 [pro] is a high-performance image generation and editing model developed by Black Forest Labs. Released in late 2025 as the successor to the FLUX.1 series, it is the flagship variant of the FLUX.2 family, specifically engineered for professional production environments and commercial reliability. The model is designed to provide a significant leap in visual fidelity, emphasizing photorealistic detail and improved spatial logic for complex compositions.

The architecture is built on a 32-billion parameter latent flow matching transformer that integrates a Mistral-3 24B vision-language model with a rectified flow transformer. This combination allows the model to process complex instructions with deep semantic understanding and world knowledge, resulting in superior prompt adherence compared to earlier architectures. By utilizing a single text encoder—Mistral Small 3.1—the pipeline simplifies the embedding process while maintaining high performance across diverse creative tasks.

Key capabilities include the ability to generate images at native resolutions up to 4 megapixels and a sophisticated multi-reference system that can process up to 10 input images simultaneously. This enables precise character consistency, style transfer, and localized image editing without the need for additional fine-tuning. The model also features significantly improved typography rendering, making it suitable for creating complex infographics and UI mockups that require accurate text and hex-code color precision.

For optimal results, official guidelines recommend using concise but highly descriptive prompts, prioritizing important details at the beginning of the text. The model responds effectively to positive descriptions—specifying what to include rather than what to exclude—and supports the use of hex codes for exact color matching. As a production-tuned variant, FLUX.2 [pro] is optimized for simplicity, aiming to produce high-quality, "finished" outputs with minimal manual parameter adjustment.