You know, the first is really around SageMaker. So our customers have used SageMaker very expansively over the last number of years because it's the easiest way to organize your data, to build AI models and then to deploy them at scale. And we have loads of customers who do this, some examples of Perplexity and Workday and NatWest. But I would say today that most companies are starting to standardize on top of SageMaker when they're building their own AI models. But what we've seen over the last year or two is that increasingly customers want their AI and their analytics in the same space. And we have a leading array of analytics services that are really good at various jobs. So like we have Redshift, which is a scalable data warehouse.
We have all these, you know, we have Glue, which does data integration. We have all these analytics services, except people have to go to each of these services when they want to use them. And what we started hearing from customers was, 'I really wished I could use your leading tools and building blocks, but from the same place and where my AI is.' And so we realized that we needed to make this easier for all of you, and we thought about should we just build a new service and call it something different? And then we realized the center right now of a lot of these AI workloads and increasingly analytics is SageMaker. And so we built a brand new addition of SageMaker, our next generation of SageMaker, which we just announced this week, and it has two key components for you.
Like Salesforce or ServiceNow or Zendesk. And we make all that data accessible in the new SageMaker. So you have all your data in one place. And then also we've built a second component, which we think of as the, we call the SageMaker Unified Studio. So we took all the studios, which you really think of those as the abstraction tools that make it easy for you to operate across all those analytics services. So studios from SageMaker, from Bedrock, from Redshift, from Glue, from Athena, all these different studios, and put them all together in a unified studio. So now you have all the data you want to do analytics and AI together in the Lakehouse, but then all those components and all those tools are in one place. So customer's very excited about this. I am too. I think it's gonna be really a game changer for you.
And then of course, we haven't stopped working on what we used to call SageMaker, we now call SageMaker AI as part of this next edition of SageMaker. And one of the interesting innovations that we launched about a year or two ago was SageMaker HyperPods. And what SageMaker HyperPods do for you is it allows you to distribute and parallelize your workloads, your training workloads across lots of training chips or GPU chips. So it allows you to speed up when you're doing training of new workloads and new models. But it also has a feature where it's constantly monitoring the infrastructure to the last checkpoint so that when it repairs the infrastructure, it just resumes seamlessly. So for most SageMaker AI customers, they are using HyperPods for their large trading workloads.
And so we've added a couple new features in the last day or so that I think are really excited. I'm gonna cite two of them. One is we've added HyperPods task governance. So this lets you do a few things. First of all, it lets you set the priority of what workloads are or what training workloads are most important to you so that we can prioritize. So let's say that you have a certain amount of compute and you've exceeded the limit of your compute, and then if you get to some sort of limit, it will prioritize the workloads that you said are most important. And it will pause at a checkpoint, the workloads that are less important and automatically resume them again when you have the capacity to be running them all. So that's the first thing it does.
Then, say, that you have designated a certain amount of compute to different priority workloads and you're not using all that compute, maybe at off peak times of the day. What we'll actually do then is reallocate that compute to other workloads so you're not stranding any of it idle. And then the third thing that's interesting is that we had this problem inside Amazon, which is we have all these generative AI applications being built and it's pretty easy to lose track on HyperPod and what amount of resource they're consuming so that you can decide, 'Do I like what I'm spending? Do I think all these are actually providing the right return to continue to spend resources on this very, very handy feature?'
And then a second feature we just launched in SageMaker HyperPod is flexible training plans. And so flexible training plans, what's interesting is you can now specify the amount of instances you think you need to be able to run whatever, train whatever your model is, but then a timeframe that you're comfortable with a training on. Sometimes you don't have to do it right away. You're willing to do it over a period of time. And then what we will do is we will return to you an estimate on how long it's gonna take and setting up all the jobs to run the workloads. It just does it all for you. It's much easier to use. So I'm very excited about the SageMaker AI innovations as well.
Second area I was gonna talk about, so yesterday you heard me talk about Q transformation, where we were making it much easier to transform from .NET to Java or Linux to be able to transform from VMware instances to EC2 instances and to be able to transform mainframes and dissect the different pieces to go from many years in migrating a mainframe into it being just a few months. And I actually really wanted to call that feature instead of Q Transform, I wanted to call it the TDK or the technical debt killer.
And so we have, you know, Q transformation is part of Q Developer and Q Developer is the very best AI assistant for software development. And so we launched three other things that we think really will remove some of this technical debt. The first is, and they're all agents by the way, as part of this. The first is really around documentation. So there are so many times where as developers, you pick up code that you haven't worked with and you just have no idea about the code. Now you can actually ask for Q to actually give you a description of that code and it will load into a README file. So you actually get familiarity with that code pretty quickly. Same thing by the way, when you build code, a lot of times you don't wanna have to take the time to document that code.
The ability to do a unit test on a piece of code you've written. But now we've expanded it so that you can actually do unit tests on an entire end-to-end project, which will speed up the work you normally have to do. And then the last piece that we added in agent four is really around code reviews. And I cannot count how many times in my history at Amazon and in AWS that we would have some sort of operational error. And when we went and did the root cause analysis, we would often find that we just missed it in the reviews, the human reviews, of when we were actually doing code reviews. And even when we had two-person reviews, we'd sometimes miss it. And so to have an agent that in an automated way, where they're not gonna be distracted
Today being the very last full day of re:Invent, I thought I would just give you a few stats I thought were interesting. So we had about 60,000 total attendees here at re:Invent this week. We had about 150,000 total sleeping room nights with a peak of 30,000 rooms on Tuesday night. There were over 21, there were 21 hotels across the Vegas strip that participated in re:Invent, hosting meetings or people sleeping there. We had 2,000 learning sessions with 247,000 session starts. About 2,000 companies brought five people or more. We'll have about 18 to 20,000 people coming to the concert tonight. We had about 1,200 people participate in our 5K, which we've been doing for a few years.
That's a record for us. And then we had someone set up in our lounge to do November haircuts and mustache trimmings, and we had over a hundred people do that. So it was a very productive, somewhat quirky, very inventive week here. And I'd just like to say that for all of us that work at Amazon, this is one of our very, very favorite weeks of the year. We work really hard to try and build you features and services and capabilities and make it easier for you to improve your customer experiences and your businesses. And we're always heads down trying to deliver it for you, but to have a chance to spend a week with all of you and to hear what you're doing on top of the platform is very inspiring. So thanks for spending the week with us and we look forward to seeing you again next year.
