Hugging Face

Confidence 0.75 · 2 sources · last confirmed 2026-07-22

Open-source AI platform founded 2016, co-founded and led by CEO Clément “Clem” Delangue — the “GitHub for AI,” where AI builders share and download open models and datasets. As of mid-2026 it hosts almost 3 million public models and 1 million public datasets (a new repository “every 7 seconds”), serves 16–17 million AI builders, and is used by roughly half the Fortune 500. It entered the wiki twice: first as the author: of the Agentic Evaluations Workshop (Mar 2026), then promoted to a full entity on its second source, the TechCrunch Equity interview with Delangue (Jul 2026), where it is the central subject.

Why Hugging Face matters in this wiki

Hugging Face is the wiki’s clearest institutional embodiment of the open-source-ai stance and the own-vs-rent thesis (see agent-harness: “the model is what you rent, the harness is what you own”). Delangue’s account of the enterprise flow — frontier APIs for experimentation, owned/open models for production at scale once cost bites — is the platform-CEO vantage on the same argument NVIDIA’s Jensen Huang makes from the silicon/substrate side (Huang interview). Positioned in the wiki’s vocabulary, Hugging Face is a collaboration/distribution platform (models + datasets + services, “part GitHub, becoming a bit AWS”) sitting beneath the agent-harness layer — the ecosystem enterprises specialize on top of.

Products and initiatives referenced in this wiki

• The Hub — the model/dataset/Space sharing platform itself; the “GitHub for AI” core.
• Open LLM Leaderboard / LightEval / Inspect AI / GAIA 2 on the ARE environment — the open-evaluation stack showcased in the Agentic Evaluations Workshop (see ai-benchmarks).
• Reachy / Reachy Mini — Hugging Face’s open-source robots; Delangue’s evidence that robotics needs open source even more than the rest of AI (petabyte-scale video/image data; a home robot shouldn’t be “a black box controlled by a few”).
• Spring 2026 report — Hugging Face’s own download analysis finding Chinese models at ~41% of downloads, surpassing the US (cited in the Delangue interview; see open-source-ai).
• Model licensing initiatives — a license type introduced “a few years ago” to give open-weight models more use-case clarity (legal-clarity contribution to the field).

People

• Clément “Clem” Delangue — co-founder and CEO. Central subject of the Equity interview but dangling (single-source, deferred) per the person-entity second-source promotion rule — central-subject status on a first appearance does not itself trigger person promotion (precedent: Jensen Huang, Sal Khan). Promote on a second substantive Delangue-authored or Delangue-centric source.

Concepts Hugging Face touches in this wiki

• open-source-ai — the platform is the institutional anchor for the open-weight / own-vs-rent / AI-sovereignty / concentration-of-power theme.
• foundation-models — hosts and distributes open-weight foundation models (GLM 5.2, Nemotron, OpenAI’s open GPT, and millions more).
• enterprise-ai-adoption — the frontier-experiment → owned-production adoption flow; half the Fortune 500 as users.
• ai-benchmarks — the open-evals / agentic-evaluation stack (GAIA 2, LightEval, Open LLM Leaderboard).
• dynamic-capabilities — digital-scouting, balancing-digital-portfolios, navigating-innovation-ecosystems, business-model, and external-triggers per the interview’s W&W tags.

Open questions

• Financials and independence — Delangue reports no funding round in 3 years (~$400M raised), a declined Nvidia investment, and near-profitability. An independent source on Hugging Face’s revenue/business model would let the wiki corroborate the capital-efficiency claims.
• The Evox Productions copyright suit (Hugging Face with Stability and Runway) — ongoing; a resolution or filing would be a useful responsible-ai/legal-risk ingest.
• The Spring 2026 report itself — cited secondhand via the interview; ingesting the primary report would substantiate the ~41%-China-downloads figure and the broader open-vs-closed trend data.