Black Forest Labs发布开源图像生成模型FLUX.3 Image Klein，Diffusers首日支持
2026年10月15日，Black Forest Labs发布开源图像生成模型FLUX.3 Image Klein，并在发布首日上线Diffusers。该模型以统一架构同时支持文生图与多参考图生图，还新增了布局控制能力：用户可通过在提示词中写入JSON格式的边界框，指定图像各部分的位置；在编辑场景中，也可通过src_bbox与tgt_bbox将参考图像中的内容移动到输出图像的目标区域。
此次发布提供两个开源检查点：蒸馏版（black-forest-labs/flux-3-image-klein）默认16步、无需CFG，适合快速推理；基础版（black-forest-labs/flux-3-image-klein-base）默认50步并启用CFG，更适合微调。据官方介绍，蒸馏版在单张NVIDIA H100上生成1024×1024图像约需2秒。模型同时提供NVFP4与FP8两种量化权重，可通过from_single_file加载以适配显存较小的GPU，并配套了面向文生图和图像编辑的DreamBooth LoRA训练脚本。
技术层面，FLUX.3采用潜空间流匹配架构，文本编码器为Qwen3-VL，并读取该模型的8层隐藏状态进行特征堆叠；图像使用与Flux.2相同的VAE，经8倍压缩与2×2 patch打包生成token。模型采用两阶段transformer结构：各模态流先独立处理，再合并进行联合注意力计算。FLUX.3以模块化Diffusers流水线形式发布，支持查看工作流、拆分组件或插入自定义块。官方还提供了Hugging Face Space，用户可在线试玩。
- 2026-10-16 02:41 | X：New from @bfl_ai on OpenRouter today: - FLUX 3 Image Klein, a new open-weights image model that run...阅读原文
New from @bfl_ai on OpenRouter today:- FLUX 3 Image Klein, a new open-weights image model that runs on one consumer GPU- An improved FLUX Video Upscale in early access, turns any video into 4K for $0.28/s (was $0.59/s)- FLUX 3 Video can now create 4K videos from a prompt
- 2026-10-15 23:50 | venturebeat.com：Black Forest Labs opens FLUX 3 Image Klein weights, with free commercial use limited to smaller businesses阅读原文
German AI video and image company Black Forest Labs is today releasing open weights for FLUX 3 Image Klein, a 5.6-billion-parameter image model designed to let developers generate, edit and customize visuals on their own hardware.The company says Klein retains capabilities from its larger FLUX 3 Ima...
- 2026-10-15 08:00 | Huggingface：Welcome FLUX.3 Image Klein in Diffusers阅读原文
Black Forest Labs FLUX.3 Image Klein is here, and it runs in Diffusers from day one. One model does text-to-image and multi-reference image-to-image, and it adds something new, the layout control. You write bounding boxes into the prompt to say where each part of the image should go, or which parts ...