Hugging Face
Hugging Face is an AI company that runs a machine-learning collaboration platform and maintains widely used open-source tools. Its website positions it as “The AI community building the future” — “the platform where the machine learning community collaborates on models, datasets, and applications.” According to public references, the company was founded in New York in 2016 by Clément Delangue, Julien Chaumond, and Thomas Wolf; it began as a chatbot app and pivoted to machine learning after open-sourcing its model.
Platform and open-source tools
The business has three layers: the Hub that hosts community resources, open-source libraries built around it, and paid services for teams and enterprises.
- Hub: as of September 2026, the official homepage counts 2M+ models, 500k+ datasets, and 1M+ Spaces applications across text, image, video, audio, and 3D. Public models, datasets, and apps can be hosted and shared for free.
- Open-source libraries: these include Transformers, Diffusers, Datasets, Tokenizers, TRL, PEFT, Accelerate, Transformers.js, smolagents, and Text Generation Inference. The Transformers documentation describes it as “the model-definition framework” for state-of-the-art models in text, vision, audio, video, and multimodal domains.
- Paid and enterprise services: Team & Enterprise plans, Inference Providers (one unified API across third-party inference services), Inference Endpoints for GPU deployment, and the PRO subscription, all reachable from the homepage.
- Robotics: according to public references, the company acquired the robotics startup Pollen Robotics in April 2025. The small biped robot MicroDuck was developed by Pollen Robotics.
Discussion in the show
Weekly #001’s chapter “Microduck 为什么一夜走红” (“Why MicroDuck went viral overnight”) discusses Hugging Face’s MicroDuck robot. Yang Pan argues that “it went viral because it was released by Hugging Face — try anyone else”, while Guizang points out it was made by a company Hugging Face had acquired, whose earlier product was already popular before the acquisition. The hosts also debate its price and its “useless enough to be fun” toy positioning.
In the business-news segment (chapter-25), Yang Pan lists “Nvidia acquired Hugging Face” among the week’s headlines; in the chapter on the value of Hugging Face and why Nvidia underpins the industry, Guizang argues that the rise of Chinese open-source models raised the value of open-source infrastructure such as Hugging Face and OpenRouter, and Yang Pan reads both acquisitions as a signal that the industry has reached a new stage. The deal itself is covered under the NVIDIA entry and the sources below; the show’s views are host interpretations, not company statements.
Frequently asked questions
What is Hugging Face?
Hugging Face is the platform where the machine learning community collaborates on models, datasets, and applications, and it maintains open-source libraries such as Transformers; see the official site. It is neither a single model nor only a model downloader.
Is Hugging Face free?
Public resources are free: the official site states that unlimited public models, datasets, and Spaces can be hosted and shared. Paid parts are the Team & Enterprise plans, the PRO subscription, and GPU inference deployment; current terms are on the official site.
What is Transformers?
Transformers is an open-source library maintained by Hugging Face. The official documentation describes it as “the model-definition framework” for state-of-the-art models in text, vision, audio, video, and multimodal domains, for both inference and training, and as the pivot that other training and inference frameworks share.
Can I download models from Hugging Face?
Yes. Public model pages on the Hub provide weights, model cards, and usage examples that work with libraries like Transformers; each model’s license is set by its author. Start from the models page.
Did Nvidia acquire Hugging Face?
According to media reports in August 2026 (reported by The Information and carried by Reuters), Nvidia agreed to acquire Hugging Face; as of those reports the deal had not closed, and later developments should be checked against official company announcements. See the NVIDIA entry for context.