In a significant move, Hugging Face has launched HUGS (Hugging Face Generative AI Services), a direct competitor to Nvidia's NIM (Nvidia Inference Microservices). HUGS provides optimized, zero-configuration inference microservices for generative AI models, making it easier and faster for developers to build AI applications.

The key differentiator between HUGS and NIM lies in pricing. While NIM charges $1 per hour per GPU, along with the Nvidia AI Enterprise suite license fee, HUGS charges $1 per hour per container. This cost advantage is expected to attract enterprises looking to deploy generative AI models without breaking the bank.

HUGS supports 13 open models, including Llama-3.1-8B-Instruct and Hermes-3-Llama-3.1-8B, with plans to add more models soon. Although NIM supports a broader range of models, including proprietary ones, HUGS' focus on open models and lower costs may appeal to developers and enterprises alike.

HUGS can be deployed on AWS, Google Cloud, and DigitalOcean, with a 5-day free trial period available for AWS users. Hugging Face's Enterprise Hub also offers access to HUGS, along with other AI models and tools, for a monthly fee of $20 per user.

This new offering from Hugging Face is expected to shake up the generative AI landscape, providing a more affordable and accessible alternative to NIM. As the AI ecosystem continues to evolve, HUGS is poised to play a significant role in democratizing access to generative AI models and accelerating innovation in the field.