Black Forest Labs’ Post

We've supercharged FLUX.1 Kontext through our collaboration with NVIDIA. 𝐊𝐞𝐲 𝐩𝐞𝐫𝐟𝐨𝐫𝐦𝐚𝐧𝐜𝐞 𝐢𝐦𝐩𝐫𝐨𝐯𝐞𝐦𝐞𝐧𝐭𝐬: - 2x faster inference on RTX GPUs with TensorRT optimization - Memory requirements reduced from 24GB to just 7GB (FP4 variant) - Full quality preservation across BF16, FP8, and FP4 quantization levels - Production-ready weights available on Hugging Face 𝗪𝗵𝘆 𝘁𝗵𝗶𝘀 𝗺𝗮𝘁𝘁𝗲𝗿𝘀:  Text-guided image editing is now instant. Edit images with simple text prompts - change colours, modify objects, update text, or transform entire scenes - all running 2x faster with 70% less memory. No more waiting. No more hardware barriers. Professional-grade image editing is now accessible on consumer RTX GPUs. https://lnkd.in/e_9G7fsE