Cofounder and CEO, Huggingface
Fireside Chat with Datadog CEO Olivier Pomel and Hugging Face CEO Clément Delangue
In the same way that open source radically changed how software is developed, openly shared Machine Learning models have fueled the rapid growth of AI. Join Datadog co-founder and CEO Olivier Pomel as he hosts Clément Delangue, co-founder and CEO of Hugging Face, the leading open source platform and hub for sharing models and datasets. They’ll discuss how AI is evolving, how organizations are responding, and what this means for you and your engineering team.
This session will be part of our livestream programming.
About Clement Delangue
Clement Delangue, cofounder and CEO of Hugging Face, made a series of media appearances in mid-2026 discussing open-source AI, regulation, and robotics. On the Equity Podcast, Delangue said he has repeatedly observed companies starting with proprietary frontier APIs and later switching to open-source models for cost reasons. He described the potential risk of a few companies controlling AI, stating that keeping models behind closed doors creates an "asymmetry of power" and argued that the world can be made safer by "leveling up the playing field" through open-source competition. Commenting on government scrutiny of Anthropic's Mythos model, Delangue said that being labeled "too dangerous" can serve as effective marketing for frontier AI firms and noted that it is "fair for the US government to at least try to get more transparency." He cautioned against spreading regulatory constraints to smaller entities such as startups and academia, which lack the resources of large AI companies. Delangue also promoted open-source approaches in robotics, citing Hugging Face's release of a small robot called "Rich Cimini" and stating that transparency and control over robots that interact with people require open-source development.
Transcript (82 segments)
I
Interviewer0:00
All right. Thank you. And I'm very pleased to have with me Clement Delangue from Hugging Face. Clem, you go by Clem, right?
C
Clement Delangue0:06
Clem. Clem is perfect. Yeah. I've been in the US for 15 years now. So I've heard any versions of the pronunciation of my name, as you know, as a fellow French person. Yes. So Clem is usually the easiest.
I
Interviewer0:19
Yes. And I apologize, you know, this is the part of the day where you get stuck in front of two guys who speak with a French accent, I hope. So Clem, I mean, I think everybody probably knows Hugging Face in the room, but can you give us a quick intro? Mostly I think the founding story is pretty interesting and how you got to where you are with the twists and turns. It's pretty original.
C
Clement Delangue0:44
Yeah. Yeah. And you've had kind of like a front row seat to that as being one of our earliest angel investors and supporters. We started really out of a passion, out of an excitement for AI, now almost 10 years ago. At the time it wasn't really called AI. People were talking more about computer vision, about machine learning, about deep learning. It wasn't as hot and sexy as it is today. And we really started with something completely different than what we do now. We started with conversational AI, kind of like a ChatGPT before ChatGPT, which obviously wasn't working at all with the technologies of the time. And after three, four years came out a model from Google called BERT, and one of my co-founders Thomas was like, 'Okay, BERT is cool but it's in TensorFlow. I want to spend the weekend putting it to PyTorch for the community.' And we did that. When we released it a few days later, it blew up online, and that's when we realized that we can not only be conversational AI but be more kind of like a platform for people and companies to use open source models.
I
Interviewer2:00
Cool. Could you give us a sense of the scale of the company and the user base, your place in the ecosystem?
C
Clement Delangue2:07
Yeah, so we're now the biggest platform for AI builders to use open-source models, open data sets. We have 15 million AI builders using the platform today. They're creating a new repository on Hugging Face. So a repository is a model or a data set, every seven seconds now. And there's like three million open public models on the platform, over a million open data sets on the platform. So it became really the central place for people and companies to build their own models, especially when they don't want to use APIs for example.
I
Interviewer2:51
Cool. And people can run the models also on your platform.
C
Clement Delangue2:54
Yeah. Usually either using our compute or integrated with other platforms and other partners. We really have kind of like the open-source philosophy that our platform is built in a way that it integrates with the whole ecosystem and also with your internal systems and internal stack. So if you want to use part of the Hugging Face platform, connect it to your internal stack or to someone else's stack, it's quite easy to do.
I
Interviewer3:25
Cool. So I very distinctly remember the first time we met. I think it was at a data meetup in New York, in late 2019 or early 2020. And you made that super convincing pitch around the fact that language was solved and that absolutely everything would be disrupted by AI and AI was taking over. And to put that in perspective, it was before ChatGPT, it was even before GPT-3. I think it was after GPT-2.
C
Clement Delangue4:02
Yeah.
I
Interviewer4:02
So of course, since then pretty much everything you said at the time came true. So what are you seeing now? What is going to come true in the future? Can you educate us a little bit? The billion-dollar question, right, for the next steps.
C
Clement Delangue4:20
What we're seeing on the Hugging Face platform, which is sometimes kind of like a sample of the future, because we have the most expert and nerdy AI builders using our platform, is a lot of traction and interest for everything except text now. We see a lot of usage on biology, chemistry, time series, robotics. So I think AI is starting to permeate really all the other parts of the economy. I'm super excited about it because thinking about AI for biology and how AI can help on drug design, on DNA analysis. We released a model two weeks ago that allows you to analyze your DNA on device, on your laptop. So these are the things we're going to talk about much more in the next few months. The second thing we're seeing which is a bit counterintuitive: we're seeing more and more smaller, more customized, specialized models. Everyone talks about these gigantic models behind APIs, but I think we're going to see in the next few months that sometimes using smaller, more specialized models is a better option.
I
Interviewer5:54
Cool. I think we can come back to some of that. I think there's a lot of questions around how each of us should be thinking about those models, whether we should adopt the big ones, the small ones, build a smaller one, and what the path from here to there can be. So if we switch gears a little bit, you run one of the largest open source platforms basically. I mean there's GitHub for code and there's Hugging Face for AI, that's more or less the split. And there's a lot of talk about what it means to be open source for AI. There's open weights, there's open code, there's open data sets. Is there a right way or wrong way to think about it? How do you like to think about it?
C
Clement Delangue6:37
No, I like to think about it as a gradient. You have pure closed source black box API, where you send an input, you get an output. You don't really know anything that's happening in between. Could even be a human interacting in the middle. You have no idea. And you have the pure open source where you have open source data sets, open weights, the scripts to create the models, train the models, run the models being open. Most of the workloads happen kind of like in between. And the more we can push the gradient towards openness, the easier it gets for everyone to participate in the creation of AI and not just a few companies. The more it creates transparency for everyone to understand what's happening, why it's happening, and the more control you give to every company, every individual, every type of organization. Not just big corporations but also little tech, nonprofits, academia of course, and a lot of other organizations. So this is a gradient and hopefully as a society, as an economy, the more we push towards openness, hopefully the better the outcomes.
I
Interviewer8:08
And you do push towards open, right? So you have a research team, you build models, you publish models, and you publish the whole thing like the data sets, the training, and everything else, right?
C
Clement Delangue8:18
Yeah, yeah. We did for example something called SmolLM, which is very competitive on the Pareto curve of small language models, where we published not only the weights as you mentioned but also the data sets and the full training scripts, so that anyone can basically retrain these models from scratch if they want to.
I
Interviewer8:44
And what do you see the rest of the industry doing there? There's been a lot of debate over the past few years around the big labs not publishing anymore. Whether that's, you know, the OpenAIs, Anthropic, Google, word on the street is if it's published, it means it's not interesting anymore, like the good stuff they keep for themselves for competitive reasons. Do you see that happen?
C
Clement Delangue9:09
We see that. When we started Hugging Face, the whole field in the US was very open with open research, open source weights, and that's very much what propelled the sector in the US. If you think of ChatGPT, the T in ChatGPT is actually Transformers, which was an open source contribution by Google. So this kind of emulation and collaboration of the whole field, in my opinion, is what created the American leadership for the past few years. It kind of stopped or regressed with a lot of these organizations being more closed source, being more secret, focusing more on monetization than open sharing. We've seen the opposite in China, where they've started to be more open, share more of the research, share more of the models, which led to this kind of weird point in history today where a lot of the best open-source models are actually coming from China and getting a lot of usage in the US. So we see a lot of companies, including enterprises, now using open weights models from China in the US.
I
Interviewer10:25
And on your platform you see the most reuse and forking and downloads for Chinese models basically.
C
Clement Delangue10:30
Yeah. End of last year, for the first time, the volume of downloads of Chinese open weight models has been higher than the volume of downloads for American open weights models.
I
Interviewer10:45
And do you see a turn there? There's a few new AI labs, whether that's in Europe or in the US, that are at least planning to release open models.
C
Clement Delangue10:54
Yeah, I hope so. We have an administration in the US now that is really in favor of open weights, open data sets. The AI Action Act from I think it was a year ago, open data sets and open weights was the third priority of the White House guidelines. And you see some organizations doing more. OpenAI released some open weights a year ago. You have an organization like Nvidia that is now the biggest contributor to open source in the US, pushing the field. We announced actually today collaboration with a company called ARC, which is a smaller AI lab that is sharing open weight models. So I think the US could get back to their roots and get back their leadership. It's going to take some time, but I think it's possible and it's something we should fight for, because that would be a great thing for the US and for the world.
I
Interviewer12:00
Yeah. And I think for everybody in the room who got to a point where they tried to customize their own models, a few years ago you would start from Llama, and today you mostly start from Qwen, or one of...
C
Clement Delangue12:13
Yeah. You see tons of examples like Cursor. They announced I think three months ago that they were post-training their own models, and a company like Cursor, a darling of Silicon Valley, funded in the US, you would think that they would use American open weights, but they disclosed that they were actually basing it on Chinese open source. And I see not even dozens, hundreds of these examples. When I talk to people, they usually don't disclose it because it might not be the most popular thing to disclose in the US, but the volume of usage of Chinese open weights in the US is really surprising. Not only in companies but also in academia. Now if you go to Stanford, to Harvard, the only way to study the state-of-the-art in AI is to look at Chinese open weights, which is kind of weird, but that is where we are today.
I
Interviewer13:20
And some labs, I know DeepSeek is publishing a lot also, so there's a lot of information that comes out there. Great. So this gets us to the idea of building with AI, not just consuming it. And I've heard you convey the message that it's important for all companies to be builders and consumers of AI.
C
Clement Delangue13:42
Yeah.
I
Interviewer13:43
Can you tell us a bit more on that? What does that mean? Can everyone build?
C
Clement Delangue13:48
Yeah. So I think the way to approach AI and to look at AI is kind of like as the new way to build software. It's kind of like this analogy from Karpathy of software 2.0. And I think what we've learned with software is that if you're a company trying to play a role in technology, you have to build the capabilities to write software yourself. You can't outsource all your software to someone else. And I think the same goes for AI. If you're using only an API, sending your data, sending your queries to an API provider and actually not building anything yourself, long term, I really think you're going to be at a disadvantage. First, this provider is going to try to take advantage of you and kind of like milk you as much as possible with their margins. Sometimes by building themselves competing products to what you actually build and pricing it to a level where it's impossible for you to compete. And second, you're going to end up with a model where it's way too costly, way too slow, and it's just going to be a struggle for you to actually build your competitive advantage.
I
Interviewer15:14
Well, that's uplifting. So everybody should build, otherwise we'll be trapped forever. No, but seriously, I think you make a great point. I think what everybody needs to realize is it's an investment, because mostly initially it's not going to pay off immediately.
C
Clement Delangue15:29
Yeah.
I
Interviewer15:30
It takes time to get those technologies to work, and by the time you customize a model, a new commercial one comes out, so it doesn't look like you've moved much. But you're in touch with the market and then you can learn and improve.
C
Clement Delangue15:47
Basically similar to software, right? If you weren't writing software before, the transition to software wasn't always easy. At the beginning maybe people thought, 'I'm not a software company, I don't need to write software myself,' but they learned, they built the capabilities, and progressively they saw the value in it. It's the same thing for AI in my opinion.
I
Interviewer16:10
Right. So another thing you have a great viewpoint into is the various trends in adoption. And in particular, it's very interesting to see right now how various parts of the world react to the rise of AI. We've all seen those polls that say that AI is extremely unpopular in the US, even more unpopular in parts of Europe, actually quite popular in Brazil, and very popular in China. What do you see in terms of adoption on your platform? Where are people downloading the models the most and building the most?
C
Clement Delangue16:45
In terms of builders, obviously the US has a great culture of building with AI, so that's where we see the most usage. Obviously a lot of usage in China, even if we're banned from China as an American company, but we still see a lot of usage from there. But we definitely also see this trend of people being scared about AI, having negative opinions about AI. To be honest, I think part of the problem comes from the AI leadership that sometimes talks and emphasizes too much the wrong issues. They don't talk enough about how AI can create jobs, how AI can solve new challenges, how AI can change the distribution of power and wealth, and instead tend to talk about catastrophic scenarios for AI, like Robocop, where AI is going to take over the world, how AI could destroy every single job that ever existed. And so that created this weird perception in the public.
I
Interviewer18:02
But you see the builders are still fairly separated from that?
C
Clement Delangue18:06
Yeah, basically. And I even see the correlation, meaning that if you give people more tools to build themselves, I think they're going to see AI for what it is, which is the new way of building software, a new tool that they can use to solve their problems. And I actually think that it's going to help on their perception. And we see the divide a little bit, right? In the US, a lot of AI is closed source, bad perception of AI. In China, a lot of it is open source, and actually you have a better perception. It's obviously not only correlated to that, there are a lot of other factors, but I think it plays a role.
I
Interviewer18:40
I see. So talking about Robocop scenarios, you're getting into robotics too but in a good way, right?
C
Clement Delangue18:46
Yeah, not a Robocop kind of thing. It's not going to kill you. It's not a giant humanoid robot that is going to do a forceful movement and kick a kid in the stomach. I don't know if you've seen this video online.
I
Interviewer19:00
I haven't seen that one, but it sounds pretty horrible.
C
Clement Delangue19:02
Yeah, it kind of looks funny at the same time, but if I were the parents of the kid, I wouldn't be happy about it. No, I mean I think it was a year and a half ago, we started to see more and more people talking to us and being like, 'Okay, I'm an AI builder, I'm building with robotics, but I don't really have the hardware to experiment with it, because it's very expensive.' And so we acquired this company back in France called Pollen that had been doing robots for seven years when we acquired them. And now we're building this little robot called Reachy Mini, which is super cute, that only costs $500 and it's kind of like an open source platform for robotics, meaning that you can program it yourself. It's fully open source, and so a lot of people use it to experiment with robotics. And it opens a window into new tasks, new use cases, new capabilities. People are using it, putting it on their kitchen counter to help them cook. Some people give it as an interface to their kid, without the kicking in the stomach, for them to interact with AI without having to look at the screen, and to learn a new language, new topic, things like that. So it's kind of like a fun extension of the use cases, and it's an example of what I was saying, meaning that AI is really permeating from only being a text assistant kind of thing to really now getting into robotics, into the real world, into industry, into biology, chemistry, and all these other domains.
I
Interviewer20:46
And that's really exciting. In a way we live in science fiction now. We talk to the computer, the computer answers. And we live in a world where more and more people start experimenting with robotics. I mean when I was growing up it was really not like that. So despite all the doom and gloom that we can read, I think that part is pretty exciting.
C
Clement Delangue21:08
My dream is that in the next few years you have like a hundred million AI builders being able to really build for their own problems, for their own challenges. And I think that would be a really great world if we can enable and empower more people to be AI builders, because it means that we won't outsource that to just a couple of labs in Silicon Valley. I think the world where you only have a few companies in Silicon Valley building AI for the rest of us is kind of scary when it comes to concentration of power, concentration of wealth, concentration of capabilities. And so if you can change that with open source and make more people become AI builders, I think we'll be in a way better world than without it.
I
Interviewer22:04
Thanks. Sounds like it's a popular idea. Have you thought of running for office maybe?
C
Clement Delangue22:10
No, no, no. An AI would be better than me at running some countries.
I
Interviewer22:15
Well, maybe in a few years, give it time. Okay. So you see also everything that's changing in AI, things are changing so fast. I literally spend 45 minutes every day catching up with what happened since the last day. What do you see, are there big changes that you didn't anticipate like one or two years ago?
C
Clement Delangue22:38
Yeah. One thing that we're seeing quite clearly is really a multi-model paradigm for companies. It's not about just using one model for everything. You're going to use one big reasoning model for some tasks, you're going to use a medium model that maybe you post-train and optimize yourself for other tasks, and then you're going to use a small customized specialized model for other tasks. So really the multi-model approach seems to be the winning one, especially with costs going crazy now. People realize that just using a big model for everything is like using a Formula 1 car to go to work. It just doesn't make sense. There's this study that was released a few months ago from Stanford that shows that 71% of the daily queries of an AI assistant like ChatGPT, 71% of them could be accurately answered by a local model, a small model that runs on your laptop for free. And what a lot of people used to be doing is sending these queries, 71% of the queries, to these gigantic models, and then they're surprised, 'Oh, my AI budget, my token budget is crazy.' Of course, if you send every simple query to the biggest possible, most expensive model, of course your bill is going to explode. So we're seeing that completely change, especially in the past few months.
I
Interviewer24:28
I like your metaphor with the Formula 1 to go to work. I have a similar one which is that right now we just invented flight and every time we want to send a postcard we buy a business class seat for it.
C
Clement Delangue24:39
Yeah, that's true, good one.
I
Interviewer24:40
And hopefully at some point we'll get some more differentiation on what we can do and also we'll be able to ship other things than postcards. Cool. So what about running models locally? Do you see that growing on Hugging Face? How important is that today?
C
Clement Delangue25:03
Yeah, we see that growing. We acquired a company earlier this year called llama.cpp which is kind of like the most popular inference runtime for local models. And we see that really as an answer to the rising costs, because obviously if you're running the model on your laptop or on the hardware that you already own as a company, it can run at a fraction of the cost of what it would be to send it to the cloud, especially if you send it to a frontier model. We see it also as a way to solve some privacy challenges. There was the WWDC from Apple, I think it was yesterday. Obviously they're leaning on this idea that if you run some of these AI models on your laptop or on your phone, it's much more private by default, by design, because you don't send your data anywhere. And as soon as you send your data anywhere, of course, you don't know where it's going to end up. So I'm quite excited about it. It's never going to be 100% of the workloads, but if it can end up being 10%, 20% of the total AI workloads, first you reduce your macroeconomic need for energy and compute, and second, I think you build a system that is more private and cheaper, more affordable for everyone to use.
I
Interviewer26:41
Yeah. And I have to say I downloaded the four billion parameters model on my phone and actually it works really well.
C
Clement Delangue26:50
Yeah. People now are starting to think more about these ratios of intelligence per dollar spent, and intelligence per watt consumed. And I think it's a great sign that the AI industry is maturing and starting to have a more nuanced view on what it means to not only do AI but make AI useful and sustainable.
I
Interviewer27:22
Yes. And speaking of efficiency per watt and things like that, there's a huge crunch in compute right now.
C
Clement Delangue27:31
Yeah, it's true. We read it in the news all the time about the AI labs having to build data centers and buy data centers at a very high price. And we also see it in the clouds. The clouds are much less elastic than they've been in the past. It's more difficult to get GPUs, it's more difficult to get access to all that compute.
I
Interviewer27:58
Do you see that change the behaviors of the builders and the model makers?
C
Clement Delangue28:01
Yeah, definitely. I think that's what's leading some of these conversations about smaller, more efficient models. The interesting thing is that we talked a little bit about US versus China. You see this problem a little bit less in China which is more open because they share their models more, and so in a way they mutualize their compute much more. Versus in the US, it feels like every single frontier lab is redoing the same training runs in silos and replicating capabilities that could be shared with more openness and transparency. But yeah, we're seeing really that as a trend for people to think and look at smaller, more efficient models.
I
Interviewer28:48
And you do let your customers or your users run models on your platform. Do you have to fight for capacity too to run those models?
C
Clement Delangue28:55
Yes, a little bit. I think everyone does. But what's pretty useful for us is coming from the background we're coming from for smaller, more efficient models. I think we worked a lot over the years on efficiency. So for example, we're seeing a lot of traction on Hugging Face from companies storing not only their public models and data sets but also their private models and data sets, a huge volume. I think last week we added three petabytes of private models and data sets to Hugging Face from companies using us just as their AWS S3 alternative. But because we've worked a lot on efficiency, we have this technology called ZIT that is basically doing the deduplication of models and data sets. So for example, when you have two data sets that share 20% of the same data, we cut that. And so it leads us to be able to do the same storage as AWS but three times more efficiently, three times cheaper.
I
Interviewer30:10
I see.
C
Clement Delangue30:10
And same goes for compute. So we can run more efficient compute, more efficient storage. And so hopefully the fight to get more capacity is not as significant as for other players.
I
Interviewer30:27
Yes. Because as you know, when you see there's a new 400 billion parameter model and you think, 'Cool, I'm going to download it,' and then you realize it's actually big.
C
Clement Delangue30:36
Yeah, very big.
I
Interviewer30:38
So, let's switch gears a little bit and look more at how you're building and organizing the company and what might be transferable to some of the people in the room that are starting to build with AI. So you run both a research and a development team. Should everyone in the room run a research team or does that just apply to you and a few other companies? What do you think about it?
C
Clement Delangue31:04
Depends how you define research team. I think every company in the world now should have a team that is learning how to run their own models, fine-tune their own models, post-train their own models. It's become really easy to do with open-source bases. You don't need billions of dollars of compute to do that. Just with a few thousand dollars, you can take an open source base and post-train it, fine-tune it on your specific task, your specific industry, your specific domain. Even more so with agents now. It's becoming much easier to leverage agents that are going to help you post-train, fine-tune your own models. So yeah, I think every single company in the world should have teams that are able to train, fine-tune, optimize, run their own models. Otherwise, as I was saying, you're kind of at the mercy of the big API providers, not only controlling everything but also raising the prices on you, which puts you in a very dangerous situation.
I
Interviewer32:13
Are there any good starting points for people in the room to start doing that?
C
Clement Delangue32:18
Oh, Hugging Face. Of course. You can really quickly find a model in your domain. If you're in banking, if you're in legal, you can look for a base model or a small model that is already customized for that. And then really quickly use some of our tools or other tools to train that very easily.
I
Interviewer32:42
Are you using coding agents internally?
C
Clement Delangue32:44
Yeah.
I
Interviewer32:44
How heavily are you using them? Is the whole team on them? You were started before coding agents actually became a thing. I think most people in the room are, maybe there's a handful of companies that were completely post coding agents. How's that going, the transformation? What does it look like for you?
C
Clement Delangue33:03
It's going well. I mean we're using obviously off-the-shelf APIs and agents, but for us we've really started to see a peak of productivity when we customized them to fit our needs more. So we have our own internal agent called Moonbot, which not only serves as a coding assistant and an assistant for the team, but also as a collaboration assistant, which is something that I think not a lot of companies are doing yet. Meaning that for example in our Slack, whenever anyone asks a question or starts a discussion, the first one to contribute is actually our agent. And that acts as some sort of knowledge mediator between team members. Because something we realized at the early stage of this process of using these tools is that at some point the productivity bottleneck switched from writing code and building product to people talking to each other. So you had kind of like the human bottleneck, where the bottleneck is not so much building stuff but people asking questions that they could have answered differently on Slack and creating distractions. Or building something completely out of your roadmap because they can, and creating these distractions for companies. So now even internal communication is somehow mediated with agents, which is pretty cool.
I
Interviewer34:47
Do you structure teams differently with agents?
C
Clement Delangue34:51
Not so much. We're already very tiny independent teams with a lot of freedom and a lot of agency. So I think that played out well in the agentic world. So it stayed quite similar.
I
Interviewer35:09
And two questions about coding agents that are more almost religious questions. The first one is do you still have to read the code in the end? And the second one is do you type or do you speak to the agent?
C
Clement Delangue35:22
We mostly type, just because we're a very distributed company and so we have a written culture that is quite strong. So even when the input is voice, we always go through the step of speech to text in between. And no, of course not, sometimes if your code is well written it can be reviewed by agents directly without a human in the loop. It depends on the criticality of the projects. We have the full range of projects where for the very experimental topics we can be quite extreme in the way we work and the way we let agents have agency. But then on the core of our platform, obviously with 15 million users, you don't want to risk an agent going rogue and breaking something. So you have to have more human oversight for the core of our platform.
I
Interviewer36:30
And do you see, so you talk to most of the companies that are building AI, do you see a lot of variability in the patterns that are being adopted there? How the companies are building, how much agents they're using, or do you feel like it's kind of all the same right now?
C
Clement Delangue36:45
It's kind of all the same because I think people don't, I don't know if they don't think too much, but it's an early complicated trend and in these technology cycles, I think people just adopt what everyone else is adopting. Which is kind of like why Frontier Lab APIs became so popular. It's almost like you don't get fired for using a frontier lab API. It's the easy early choice. But I see that progressively changing with people maturing and feeling more confident about doing and building what they think is right and not just what everyone is doing. So now I see more variability in what people are using and how they're using it. We have like 50% of the Fortune 500 that has verified organizations on Hugging Face, probably even more than that, using Hugging Face with a verified enterprise account. And the types of models that companies are using is quite large. The long tail is quite long.
I
Interviewer38:06
So now about the way you run Hugging Face. You run both an open source platform and a product company. How do you marry the two? Is it the same people doing both sides or can you fit both into the same head, or do you have to separate a little bit?
C
Clement Delangue38:22
Yeah. It's usually kind of like the same teams. Especially with agents, I think now you can go more full stack than you used to. And you see that with the Frontier Labs. If you think of OpenAI doing a consumer B2C chatbot and an API, I mean 10 years ago an investor would tell you it's crazy. You're either an API B2B business or a consumer product, you can never or very rarely do both. But I think now it's become more common, and so most of the teams they try to do both when it makes sense and it works quite well. One peculiar way we've organized the company is of people being more generalists, being able to do more things rather than specialists, and it served us well in the past few years.
I
Interviewer39:16
Okay. Well, is there anything you want to leave the audience with? Something that they can act upon, or it's your chance to really shock the audience if you want to become a meme. The floor is yours.
C
Clement Delangue39:29
Well, I won't do the usual thing that I become a meme for, which is doing the Hugging Face emoji face like that. But I think now people have seen it too much for it to become a meme. No, my only advice is to try different kinds of models. Experiment with open source models, with smaller models. Build the internal capabilities for yourself to be AI builders and not just outsource to APIs. That's kind of the biggest shift and the biggest trend that I've seen in enterprise in the past few quarters.
I
Interviewer40:10
Well, Clem, thank you very much. I have to say it's super impressive to see the growth and the impact of Hugging Face. And I'm personally very happy to see that the ecosystem in AI has a shot at being a very healthy ecosystem with a lot of large commercial vendors, open source vendors, technology that comes from various parts of the world. So thank you for playing a part in that and thank you again.
C
Clement Delangue40:36
Thank you. Thanks everyone.