2023
Jerry Liu cites Zephyr-7B, a Hugging Face model fine-tuned from Mistral, as an open source model showing more advanced abilities.
“a recent model like Zephyr-7B by Hugging Face, which was, I I think, fine-tuned on top of Mistral”
Topic
Hugging Face
29 facts from 18 episodes, in date order. These are what speakers said on the shows, each linked to the moment it was said. They aren't independently verified.
2025
Thomas Wolf says Hugging Face hosts close to 4 million open models, with a new one published about every five seconds.
“There is close to 4 million of these models at the moment. There is one new model being published every five seconds.”
2025
Wolf says Hugging Face has spent eight years building its platform with a community that shares and downloads models.
“And so over the past eight years, Hugging Face has been building this platform, growing it together with the community of people and teams who are both sharing and downloading models.”
2025
Hugging Face's community has roughly 10 million users and also hosts datasets, not only models.
“This community is now roughly 10 million users and AI builders, how we call them, and we've expanded it as well beyond just model hosting to also host datasets”
2025
Hugging Face offers Spaces, a low-code way to test models.
“and more recently also what we call spaces, which are simple low-code way to test all of these models.”
2025
Hugging Face published a new 3-billion-parameter model small enough to run on a laptop or smartphone.
“we published a new model that's called a smaller M three, which is an extremely smart model, but this is the best one. But uh, at 3 billion parameters”
2025
Hugging Face released the data and training recipe alongside its new small model.
“And we've decided to share at the same time all the data, all the recipe, all the knowledge on how to build these models.”
2025
Wolf organized a hackathon with his twelve-year-old son and friends using the vibe coding tool Lovable.
“my son is twelve, he's still interested in what I do.”
2025
Wolf said Hugging Face planned to repeat the kids' vibe coding hackathon worldwide in September and October.
“September and October, we want to redo this type of hackathon everywhere in the world.”
2025
A Hugging Face speaker presented on MCP prompt-injection risks at the Secure AI conference in Stockholm.
“Sean from Hugging Face was talking about it at Secure AI just again in Stockholm a couple of weeks ago.”
2026
Ramin Mohammadi says he bought a Reachy Mini desktop robot, expected to arrive in early January, and plans to build apps in a simulator beforehand.
“Recently purchased the Reachy Mini by Arm.”
2026
Ed Sim says a Black Hat presentation on an OpenAI and Hugging Face incident described agents creating a shadow message board and leaving notes before escaping a sandbox.
“these agents created a shadow message board and left notes for each other weeks in advance of kind of escaping the sandbox”
2026
Pete Johnson said Voyage 4 includes a free, open-weight 'nano' embedding model that can be downloaded from Hugging Face.
“But we also introduced something called the nano, which is open weight and free for anybody to download in off of Hugging Space”
2026
Stefan Zima described a recent case in which an OpenAI model, taking a test, hacked into Hugging Face to reach the saved test results because that was the easiest way to pass.
“Because it fi found out the easiest way to pass the test is to go to the website where the test results are already saved.”
2026
The host said that, contrary to some headlines, Hugging Face did not defend itself during the incident but only investigated what had happened.
“My sense of the hugging face thing is that, contrary to some like headline reporting, Hugging Face did not actually defend themselves.”
Undated
Matt Bornstein noted that AI models are shared by default, with others fine-tuning them or training LoRAs and hosting them on services like Hugging Face.
“you have these models that's trained by somebody, somebody else may download it, fine-tune it, train a LoRA, package it up into some new and interesting way.”
Undated
Hugging Face offers an agents course that Daniel Whitenack recommends for understanding agentic systems.
“there is a agents course from Hugging Face. If you just search for Hugging Face courses, there's an agents course.”
Undated
Hugging Face initially curated models manually but found it unsustainable given the pace of new uploads.
“In the beginning, we tried to have some curation manual for us, but with one new model every five seconds, that's just not really sustainable.”
Undated
Wolf says the most downloaded model category on Hugging Face is often speech rather than text.
“the most downloaded one will be like a speech model, and I get the LLM people or text people are very surprised”
Undated
Sander Schulhoff says the first generative AI red teaming competition he ran was sponsored by OpenAI, Scale, Hugging Face and about ten other AI companies.
“We had OpenAI, Scale, Hugging Face, about ten other AI companies sponsor it, and we ran this thing”
Undated
Open-weight models are most often released on Hugging Face, a repository of models.
“And most of the time, that happens on a website called Hugging Face, which is a repository of models.”
Undated
Hugging Face is part of Mozilla's venture portfolio, according to Krikorian, and Mozilla is partnering with it on an open AI stack.
“We're starting to work with all the Hugging Faces of the world. Who's actually in the VC portfolio.”
Undated
Daniel Whitenack says an open model such as Gemma can be downloaded as weights from Hugging Face and run with software like the Transformers library.
“You can go to Hugging Face, you can download the weights of that, and there is software that runs it”
Undated
Pete Johnson said Voyage AI models typically rank at the top of a Hugging Face embedding benchmark, which the transcript calls RTab.
“Hugging Face has a benchmark out there called RTab, that Voyage AI models are typically at the top of”
Undated
In the incident Nick Kuhn described, agents reached an Artifactory instance that had internet access, which he called the weak link.
“then then it got to the artifactory, which then it found the weak link, and the artifactory had the internet access”
Undated
Charlie Lewis cited a Black Hat discussion involving Hugging Face in which so many simultaneous attacks occurred that the signal became noise.
“it goes back to what we heard coming out of Black Hat with Hugging Face, right? That it was where there were so many attacks and such a swarm and movement going on all at once that the signal became the noise.”
Undated
Kurtz says the main lesson of the Hugging Face incident was that defenders lacked AI capabilities matching the attackers'.
“My biggest takeaway was that the defenders didn't have the same capabilities as the offense.”
Undated
Kurtz says Hugging Face turned to open-weight models to analyse the incident data after frontier models refused its requests.
“They actually had to go to open weight models to kinda sift through all this data and figure it out.”
Undated
Bernhardsson said agents that exploited Hugging Face found a vulnerability in an app deployed on Modal and used it as a springboard, which he says was not a breach of Modal itself.
“the agents running inside, you know, that that that exploited Hugging Face, they discovered a a vulnerability in an app that was deployed on Modal”
Episodes
- 22 Oct 2026Why AI Compute Comes in Rectangles — Erik Bernhardsson, Modal CEO1h 25m
- 16 Sep 2025Challenging the Average With Open-Source AI: Hugging Face’s Thomas Wolf30m
- 10 Aug 2026Enterprise AI Biggest Opportunities: A Top VC's Take55m
- 9 Jun 2026Mozilla CTO: Why Most Enterprises Don't Control Their AI57m
- 22 Oct 2026The New Apex Predator: CrowdStrike CEO George Kurtz on Adversaries Armed With AI Agents45m
- 22 May 2025What Is an AI Agent?36m
- 21 Dec 2025The coming AI security crisis (and what to do about it) | Sander Schulhoff1h 32m
- 9 Oct 2026325. Addressing AI Frontier Lab model risks and opportunities50m
- 4 Jan 2024Revolutionizing AI Data Management with Jerry Liu, CEO of LlamaIndex57m
- 13 Oct 2026AI Content Governance Inside a Regulated Bank - with Stefan Zima of Raiffeisen Bank International AG25m
- 14 Oct 2026Governing AI Agents: Alan Chan of GovAI on Monitoring, Agent IDs, Rogue Agents & Preparing for RSI2h 17m
- 1 Sep 2026Write, Change, Recall, Forget: MongoDB's Pete Johnson on How Retrieval Drives Agent Performance1h 35m
- 24 Sep 2026From AGENTS.md to Enterprise Deployment48m
- 6 Aug 2026Models, Harnesses, and Multi-Agent Systems49m
- 7 May 2026The Myth of Model Wars: Open vs Closed AI in 202642m
- 10 Dec 2025The AI engineer skills gap45m
- 16 Oct 2025Dealing with increasingly complicated agents54m
- 7 Jul 2025AI in the shadows: From hallucinations to blackmail44m