NVIDIA announced optimizations for FLUX 3 Image Klein on Blackwell GPUs, delivering up to 3.46x faster text-to-image generation and up to 4.45x faster image-to-image generation compared to unoptimized BF16 inference. The optimizations are aimed at developers working on image generation and editing tasks.
What happened
NVIDIA announced optimizations for FLUX 3 Image Klein on Blackwell GPUs, delivering up to 3.46x faster text-to-image (T2I) generation and up to 4.45x faster image-to-image (I2I) generation compared to unoptimized BF16 inference. FLUX 3 Image Klein is an image editing and generation model from Black Forest Labs (BFL) trained on NVIDIA GB200 NVL72 and optimized for NVIDIA GPUs. The optimizations include NVFP4 quantization, fused PrimsTS GEMM kernels, and torch.compile, which together provide significant speedups. The optimizations are aimed at developers working on image generation and editing tasks, reducing inference latency and improving the efficiency of visual production workflows.
What to weigh
- The optimizations are specific to NVIDIA Blackwell GPUs.
- The optimizations require specific environment configurations.
Source: NVIDIA
