Hugging Face

Hugging Face, a French-American firm, offers an open-source and ML platform that allows developers and ML practitioners to collaborate and develop entire AI/ML models from training to deployment.

Hugging Face offers a range of tools for data science hosting and development. These include a platform akin to GitHub for AI code repositories, models, and datasets, along with web applications for demonstrating AI-powered applications. Additionally, it provides libraries for tasks such as dataset processing and model evaluation. It has an enterprise version of the platform that supports both software-as-a-service and on-premises deployments.

The company's paid features encompass AutoTrain, which streamlines and automates the process of training AI models; Inference API, which enables developers to host models without the need to manage underlying infrastructure; and Infinity, which is aimed at enhancing the processing speed of in-production models with data.

In addition, it also offers pre-trained NLP models like BERT, GPT-2, and GPT-3 in its repository, known as the "Model Hub," which serves as a platform for sharing, downloading, and fine-tuning pre-trained models, fostering collaboration and reusability in NLP applications.

Hugging Face acquired Seattle-based data storage startup, XetHub in August 2024 to integrate XetHub's technologies into the Hugging Face hub.

In September 2024, the company launched LightEval, a new lightweight evaluation suite to assess large language models (LLMs). LightEval provides a customizable evaluation suite that works with Hugging Face tools like Datatrove and Nanotron. It supports evaluation on various devices, including CPUs, GPUs, and TPUs, and can scale for both small and large deployments.

In October 2024, Hugging Face launched Gradio 5 to develop AI-powered web apps and improve AI adoption of ML technologies.

Key customers and partnerships

As of February 2024, Hugging Face serves over 10,000 customers across 50,000+ organizations. Its revenue mainly stems from consulting contracts with major AI companies like NVIDIA, Amazon, and Microsoft.

Hugging Face entered a strategic partnership with Google Cloud to allow developers working on Hugging Face to leverage Google Cloud's infrastructure and Amazon to make its platform available on the AWS marketplace. It also partnered with SafeCoder, a code assistant solution, to incorporate it within the Hugging Face platform to help customers develop new codes as well as improve existing code while preserving code privacy. In addition, it partnered with Graphcore to release 10 new transformer models designed for Graphcore's IPUs.