Stability AI and NVIDIA Unveil Stable Diffusion 3.5 as NIM Microservice

Details

• Stability AI has introduced Stable Diffusion 3.5 as a high-performance NIM microservice in partnership with NVIDIA, aiming to streamline enterprise deployment of advanced image generation models.
• The collaboration targets creative teams and organizations seeking faster, scalable AI solutions, leveraging NVIDIA’s NIM microservice platform for containerized, ready-to-deploy models.
• The Stable Diffusion 3.5 NIM boasts a 1.8x speed increase over PyTorch, reducing image generation times on H100 GPUs from 6,800ms to 3,700ms, and supports model variants such as Depth and Canny ControlNets within a unified container.
• This release builds on NVIDIA’s NIM ecosystem launched in March 2024, which focuses on packaging AI models and optimized inference engines into easily deployable containers for rapid enterprise adoption.
• The microservice is compatible with data center-class Ada and Blackwell GPUs and offers flexible licensing options, including a permissive community license for smaller organizations and enterprise plans for companies with more than $1 million in annual revenue, available at build.nvidia.com.

Impact

This collaboration empowers enterprises to deploy cutting-edge AI image generation with lower latency and simplified infrastructure management. The performance boost and ease of deployment could drive wider enterprise adoption and signal the industry’s shift toward containerized, microservice-based AI production environments.