--- title: 'GPT-5 Launch Reaction and Live Testing' source: 'https://youtube.com/watch?v=sIDi85HfWVA' video_id: 'sIDi85HfWVA' date: 2026-07-14 duration_sec: 0 --- # GPT-5 Launch Reaction and Live Testing > Source: [GPT-5 Launch Reaction and Live Testing](https://youtube.com/watch?v=sIDi85HfWVA) ## Summary The video is a live stream reaction to OpenAI's GPT-5 launch event. The host watches the official presentation, which includes demos of GPT-5's capabilities in coding, writing, voice, and health, as well as announcements about availability, pricing, and new API features. The host and chat discuss their impressions and test GPT-5 in real-time. ### Key Points - **Stream Start and Expectations** [00:00] — The host starts the stream early, discussing GPT-5 rumors and leaks, including a GitHub leak of model names (GPT-5, mini, nano, chat). A poll shows most viewers expect GPT-5 to be 'as expected'. - **GPT-5 Launch Begins** [14:00] — Sam Altman introduces GPT-5 as a major upgrade over GPT-4, comparing it to talking to a PhD-level expert. He highlights software on demand and on-demand expertise as key features. - **Benchmarks and Capabilities** [18:00] — Mark and Max present benchmarks: GPT-5 sets new highs on SWE-bench (coding), MMU (multimodal reasoning), and AIME 2025 (math). It also shows improved factuality and health-related performance. - **Availability and Pricing** [25:00] — GPT-5 rolls out today to free, plus, pro, and team users. Free users get GPT-5 with limits, then switch to GPT-5 mini. Pro subscribers get unlimited GPT-5 and GPT-5 Pro extended thinking. API pricing: $1.25/$10 per million tokens for GPT-5. - **Coding Demos** [30:00] — Demos show GPT-5 building a physics demo (Bernoulli effect), a French learning web app, and a 3D castle game. The model writes hundreds of lines of code in minutes, with automatic reasoning and self-correction. - **Voice and Personalization** [45:00] — Voice improvements include natural conversation, study mode, and language learning. New features: customizable chat colors, personalities (supportive, professional, sarcastic), and memory with Gmail/Google Calendar integration. - **Safety and Training Innovations** [55:00] — GPT-5 introduces 'safe completions' to handle dual-use prompts by providing partial answers and alternatives instead of outright refusal. Training used synthetic data from O3 to create a high-quality curriculum. - **Health Use Case** [65:00] — A patient shares how GPT-5 helped understand biopsy results and make treatment decisions. GPT-5 scores highest on Healthbench, a physician-created evaluation. - **API and Developer Features** [75:00] — API includes GPT-5, GPT-5 mini, and GPT-5 Nano. New features: custom tools (free-form text), structured outputs with regex/grammar, tool call preambles, and verbosity parameter. Context window expanded to 400K tokens. - **Cursor Integration and Enterprise** [90:00] — Cursor CEO demonstrates GPT-5 fixing a real bug in the OpenAI Python SDK. Enterprise examples: Amgen (drug design), BBVA (financial analysis), Oscar (clinical reasoning). 2 million US federal employees will get GPT-5. ### Conclusion GPT-5 represents a significant upgrade in reasoning, coding, and personalization, with improved safety and availability across all tiers. The host and chat find it promising but note that real-world testing will reveal its true impact. ## Transcript Okay. So, let's see here now. 29 minutes. Hello, Edgar. Nice to see you back here. pretty excited for GPT5. Hope it will be in the API from the start. Yeah. Uh me too. I um I went over there now just to check. I guess there are no early releases. Hello Scott. Nice to see you here. Stan Sten Tulu. Nice to see you. Richard pretty excited, right, about the new release. At least I am. It's been a while. When was GP4? Hello, Dimma. Mark, nice to see you. Aruni, Unreal, Link, nice to see you back. Uh, yeah, I agree. Mark, hello Narius. Sebastian, nice to see you here. pretty excited. GPT4 release. So that was over two years ago. Hello MS, Adam Pixie, nice to see you all. So GPT4 March 20 March 14th, 2023. Camille, nice to see you. R setat LV nice to see you back to Pakula hello Poland good so it's been over two years since they did like a a number upgrade that's pretty crazy hello Mark Turken yeah I'm going to live stream do like a watch along that's my fan. Uh, I nice to see you, Charlie. Uh, here are some more results from the local GP OSS. Really nice in open code with the tool finetune. Okay, I'm going to try it out tomorrow Edgar. Hello, Jess from LA. Good morning, I guess. Control plus dad plus son. Nice to see you here. Today is the launch here. Paulo Solves, hello from Latia. Thanks, Rich. I appreciate it. Uh, creative art. It's in, let me see, 26 minutes. Live a little. Nice to see you here. Yeah, I guess it's the morning on the west coast now of US. OSS doing with aentic tool use uh aentic tool usage. So I guess Edgar has been testing it out. So I'm probably going to try out tomorrow, but I'm going to be pretty busy with GT5 anyway, I guess. Hello other bio. Good to have you here. Uh Rish uh what will you be doing in this video? Um basically I'm just going to watch and do some commenting and if you want to watch on your own and people can react in the chat, right? That is basically the idea. People want to meet somewhere to watch it together. That was the idea. I asked earlier this week on stream and people wanted to do it. Hello, Chem $100 sub like Entropic. That would be cool. Cool. Right there. I think there are some rumors about this uh $200 pro subscription going to continue or something. Uh I'm in Norway creative art. Louis P, hello Christian Rosa, nice to see you. As far as local tool calling, might have something good with the 20B model. Cool. So, it's pretty crazy to think about that. It's been over two years since a new uh number update on the GPT models. I was for sure expecting it's going to be sooner than August 25 when they release GT4, but uh I guess they took their time. Hello. The $200 sub is useless right now in OPI. Yeah, I agree. 86 SEC from Hong Kong. Hello from Denmark. Unil Quest Circon. Nice to see you back. So, hello Blizzard. I thought I thought I was going to do a poll, right? What is like the most uh isn't it a humanities uh exam benchmark? Something like this. They always refer to this. Do they have like a leader bar here? I'm not quite sure. I guess so. I thought we can do like um April 23rd. Where is that benchmark? I'm not quite sure where to find it. Hello Kamal Jake. Nice to see you back. Hello G. Yeah, I'm Norwegian. Do you use cloud code for your task? Uh Cam. Yeah, lately I've been basically switching from u cursor to cloud code. So mostly cloud code. other bio. It's in let's see about 20 minutes. I thought I could just start the stream a bit earlier. Hello Vre, nice to see you back. Hi Edgar. If you can do English will be better for people in chat. But nice to see you here. Freas and Mario. Yeah, I kind of jumped the gun, didn't I? But I just wanted to to get this stream going. There's nothing else to do, is it? Can you handle using Sonnet or are you just giving it token wise and spamming opus? So I kind of do include code. I kind of do models and I do kind of I set it to set and if I'm going to do something else I'm just going to switch to opus. If I don't do that I just burn my opus tokens right away. Right. Thanks Kamal. I appreciate it. GU, nice to see you here. GPT5 API will be same price as 03. That could be good. Uh 03 API price. I always forget these pricings, right? So 03 is two in eight out. Oh, that's going to be pretty. I will be surprised if it's that cheap. All right. So, what is a benchmark they probably going to refer GPD5 to? What do you guys think? Is it going to be that humanities uh last exam? Where do I find that? It's not updated. Would be nice to would be nice to do a poll what people think the benchmark is going to be for one of the most the toughest. Do you think they're going to do Arc AGI? Arc AGI. Okay. Which version? They posted some benchmarks. They crushed the L2 uh benchmark. Is there any official rankings on that? I'm not quite sure. Archi, yeah, they're just going to do it against their own model. I agree with that. Hello Alan. Fabiana, nice to see you here. Does anyone have comp any comparisons for cloud code uh set for via the API or per use? Yeah, I'm not quite sure. Sebastian. Yeah, there are some leaks. Uh I had a look at uh like I went to Reddit now and there's like a bunch of different leaks here. All right, so here's one. GPD 5 will represent nearly 25% increase in version number of GP4. Whatever that means. It's kind of hard to uh I saw GitHub made a big mistake, right? GitHub I think they actually leaked the model names. That's pretty crazy. How did they do that? I saw this Arc 2 is pretty low in numbers. Yeah. Grop beating two 2.5 Pro at Google chess benchmark. Yeah, I s I think I saw a YouTube video on that ch. Hello Kel from Macedonia. Comparisons with Opus 4.1. I guess we're going to in the next few days we're going to get like a lot of updates. I think all leaks are intentional. Yeah, they could be that. Of course, they want to leak something to hype it up, right? Let's do I have a poll I want to run. So, let's do uh do a poll GPT5 prediction. Uh, I'm going to do disappointing as expected uh over uh surprised. So, what do people think? Do you think um it's going to be disappointing or as you expect or you will got get surprised? There's been so much leak leaks now anyway, I guess. So, uh I think it's just going to be as expected. I'll be surprised if I get really surprised by the by the stream. Curious if they can consolidate O and GPT models. GPT5 in can be used in AS. I don't know. Maybe. I'm not sure. I'm expecting GPT5 to be really smart about switching the best model for the task at hand. Hello Ken. Nice to see you. Tim, do you like choosing your model uh or omni model with no choice? It's going to be exciting to see if they're going to do that routing experiment I talked about. Is it going to be just one up in the model picker here? Is it going to be just GPT5 and maybe some other things? I'm not quite sure. Yeah, most of the prices are kind of climbing up. If you want kind of the premium uh premium plants, it's getting expensive. Hello, Christian. Nice to see you. Surprised if you're surprised. Yeah. Yeah, as expected, Tony. I kind of agree. So the leaks from uh here were the kind of leaks from GitHub. GitHub did like a post today I think. Hello Tart. Nice to see you. Ara. So this is kind of the leaks uh GitHub did. So it's going to be a GP5, a mini, a nano, and a chat. I don't know quite what they mean by chat but I guess is that going to be non-reasoning do you think maybe nice to see you here we are going to watch the open AI stream in 14 minutes it's highly anticipated 2300 thousand people waiting in the stream. Maybe chat is also for voice. Yeah, that might be it. Chat might be the m main interface. Yeah, that could be it. It's going to be interesting to see how they're going to redesign kind of the chat GPT page here. I think they're going to get rid of this model picker here. It's getting really hard to use. It's not so very user friendly anymore for new users. And I basically only use 03 though. But how to run 120 gigabyte models locally? Well, uh, Creative Art, you kind of need you kind of need, uh, some good hardware then. Can't tell you exactly how much you need, but probably a lot. Maybe 48. I'm not quite sure. Hello, Edgar. Ken, nice to see you here. I'll be so pissed if they release GP5 preview for a couple of months as always. I wouldn't be surprised. Cam. Hey Nick. They usually like to do that. Uh if you go to if you go to the playground here, they like to do the the previews, right? They usually do that. Hello night spider. Okay, so let's end this one. So pretty everyone think it's going to be around as expected. Uh, who thinks uh GPT 5 uh in API only for pro. This is kind of I hope not. Or I guess maybe it's not pro. It's like uh tiered, isn't it? I kind of wanted to try out the API today, but we'll see. Maybe not. Sami is always overhyping with safety. Yeah, just watch a video for Jeff Gearling about AI clustering. Early infant stage of the vertical scaling is the game. Hello, Andrew. Codex. So, I need to test the sound. Uh, but I kind of did like a plug in here. So when the stream starts, I can just uh kind of adjust the sound here so we get it right for most people. You think nano for free? No for free. Uh are we going to upset uh Briana if uh it was only for pro? They can't do that, right? They've been hyping this GT5 for so long now. If it's just going to be for pro rolling out over the next uh three weeks, that's going to be annoying, right? Uh Codeex. Yeah. Starting in 10 minutes. 10 minutes. Are we going to get the supernatural flirty AI now with the showcase a good while ago? David, maybe they bring back the Scarlett Johansson voice. Hey Nick. Uh, yeah, I can do that. Uh, I'm I'm kind of reworking my repo anyway now, so I can just do it. Uh, we can look at it maybe tomorrow or something. We'll figure it out. Congratulations, Chem. So, it's going to be Okay. So now, okay. So, let's test the sound now. I probably need my earphones. Yeah, I'll check it. Uh, I'm just going to test it out. Tick tock. Uh, okay. So, let's test the sound now. So, can I do like a comment here or is it too Is that too loud? Should I turn down my mic a bit? Works fine. Okay. Okay. I need to Yeah, I can't do full screen. A bit too loud. Okay, so most people think it's fine. Sounds good. Okay, that's great. We can just adjust it uh while we go anyway. Doesn't matter too much. Yeah, these live streams are a bit they're a bit strange the formatting. Let's see what happen. Hello Banks. Nice to see you. So Sam Alman confirms the live stream will be about an hour long. So they probably had a lot to talk about. Too loud. Okay, I can turn it down a bit for now. And I will put it up again when they start talking. Right. This will be a marginally disappointing for Tot. Uh Norman's brains are getting a major upgrade. Yeah, I think actually I think it's going to be overall like a win for most people anyway. Maybe they'll do a GP Pro, GPT5 Pro. Uh, I think so. I think I saw something about uh a GPT uh 5 Pro tier, right? I don't know where I saw it, but somewhere I think it was uh here. So, pro is going to be access to GD5 with pro reasoning or something, but we'll see. It's just minutes. Anyway, hello Roy Sid. Nice to see you here. Yeah, we're going to get a pro, right? They need that. AGI or bust them cool X. I don't really understand the formatting on live streams. It's a bit strange. You can't really I haven't streamed a live stream before. It's a bit different than uh uh than other videos, but I guess it's fine. What are you most excited about at GBT 5? I think I'm most excited about uh so for me like overall it's just going to be if it's really good it's just going to keep the acceleration going right if this is like a a leap we're just going to get more competition and uh I think that's what's most exciting and and of course if they can compete with anthrop ropic for coding. Uh it could be could be also good for the overall U. Yeah, I tried F11. It doesn't it's very strange. Second part of the live stream uh will be for developers. Okay. So that's probably got to be the API then. Let me just test and see if there is >> say we're able to get to that. >> Yeah, it's just a live stream because this this video goes full screen. Uh, but the live stream, uh, it's kind of like a special formatting. I don't know why. Okay, so countdown 3 minutes. I guess I can turn up the sound of it. Yeah, 1 million context window. I didn't think of that. Let's ask for that uh 1 million context plus. Who thinks it's going to be over 1 million contacts? 10 million infinite. That would be cool. No way. 1 million. So the spicy sword searches not a hope. 1 million contacts means nothing if the model can't kill off. That's true. But um for some tasks I really I sometimes only use I guess if the 1 million context window doesn't really uh if it doesn't work as intended. There's no point. If they give a larger memory for longer horizon independent task would be a game changer. Yeah I hope to see something about like uh long running tasks that would be cool. Hello Alex. Yeah, would be a bit disappointing with uh 256 or something, right? Free Sora. Don't really care. 1 minute the lights off. I do remember when 32K was uh when I started in like GFD3 beta. I think the context window was like 2K or 4K. Yeah, it's probably going to be like a bit surprised if um surprised if I'm very surprised. It's probably going to be as expected. How is the sound by the way? Should we turn it up a bit or just let me know when uh they start talking. Okay, here we go. >> Good morning. 32 months ago, we launched Chat GBT and since then it has become the default way that people use AI. In that first week, a million people tried it out and we thought that was pretty incredible. But now about 700 million people use Chat GPT every week and increasingly rely on it to work, to learn, for advice, to create, and much more. Today, finally, we're launching GPT5. >> Yay. >> GPT5 is a major upgrade over GPT4 and a significant step along our path to AGI. >> Nice. >> Now, today we're going to show you some incredible demos. We'll talk about some performance metrics, but the important point is this. We think you will love using GPT5 much more than any previous AI. It is useful. It is smart. It is fast. And it's intuitive. >> GPT3 was sort of like talking to a high school student. There were flashes of brilliance, lots of annoyance. Uh, but people started to use it and get some value out of it. With GPT4, maybe it was like talking to a college student. Real intelligence, real utility. But with GPT5, now it's like talking to an expert, a legitimate PhD level expert in anything, any area you need on demand that can help you with whatever your goals are. And we are very excited that you'll get to try this. But it's not only asking. Now, GPT5 can also do stuff for you. It can write an entire computer program from scratch to help you with whatever you'd like. And we think this idea of software on demand is going to be one of the defining characteristics of the GPT5 era. It can help you plan a party, send invitations, order supplies. It can help you understand your healthcare and make decisions on your journey. It can provide you information to learn about any topic you'd like and much more. This is an incredible superpower on demand that would have been unimaginable at any previous time in history. You get access to an entire team of PhD level experts in your pocket helping you with whatever you want to do. And anyone pretty soon will be able to do more than anyone in history could. So today we're going to talk about GPT5. We'll show you some upgrades to chat GBT and we'll talk about the API. GPT5 is great for a lot of things, but we think it's going to be an especially important moment for businesses and developers, and we're very excited to see what they're going to build with this new technology. So, we can't wait for you all to start building with this. We can't wait for you all to start building with this. We hope you enjoy it as much as we enjoy building for you. And to start, I'm going to hand it over to my colleague Mark, our chief research officer, to tell you about GPT5. Thank you. Okay. >> Hi, I'm Mark and I'm joined by Max who leads the post training team and Renie from our engineering team. Over the past few years, OpenAI has spearheaded the reasoning paradigm. These are models which pause to think before delivering more intelligent responses. Now reasoning is at the heart of our AGI program and it underlies the technology that we use to ship stuff like chatd agent and deep research. GPD5 aims to bring this breakthrough to everyone. Until now, our users have had to pick between the fast responses of standard GPTs or the slow, more thoughtful responses from our reasoning models. But GPD5, it eliminates this choice. It aims to think just the perfect amount to give you the perfect answer. Okay. >> Now, something like this takes a lot of hard work. We've had to do a lot of research to make CHPD5 the most powerful, the most smart, the fastest, the most reliable, and the most robust reasoning model that we've shipped to date. Today, we're going to show a series of demos in coding, in writing, in learning, and in health. But GPD5 isn't limited to these domains. It's very useful in all cases where you require deep reasoning or expert level knowledge in things like math, in physics, even in things like law. And the exciting thing is we're excited to make this available to everyone, even to our free tier. After we show our demos, we're going to be talking about how GBD5 supercharges our chat GPD app and our API. We believe that GPD5 is the best coding model on the market today. To start, let's head back and talk a little bit about the benchmarks and how the models stack up. >> Yeah, thanks Mark. So, as Mark said, we think GBD5 is by far our smartest model ever. So, let's start by talking through some eval. Now, eval aren't everything and they don't tell you everything about a model, but they can highlight its intelligence. And GPG5 performs exceptionally well on a range of academic evals across subjects. It outperforms both our previous models and other models on the market. So picking up first on the theme of coding, GPD5 sets a new high on SWEBench, which is an academic eval that tracks performance on real software engineering tasks. >> What do you guys think? >> Now this again is an eval, but we think it will reflect the model's performance in the real world. GPD5 also performs very well on ADER polyglot, which measures its ability to implement complex functionality in a variety of different programming languages. Now, beyond coding, GPD5 performs exceptionally well at multimodal reasoning, setting a new high on MMU, actually outperforming both our previous models and most human experts on this task. This is basically a visual reasoning domain where you are asked to from an image figure out what's going on. Uh, GBD5 is also excellent at mathematical reasoning as shown by its performance on Amy 2025. Now this is an exam that American high school to students take to qualify for the international international mathematical olympiad and GPT5 performs exceptionally well again beating our previous models and other models that are out there. >> Now moving beyond academic eval towards some real world use cases. We put a lot of work into making GBT the most reliable and accurate model in the world. Language models historically have been plagued by hallucinations, factual errors that make it hard to rely on their outputs for actually important tasks. For GBD5, we made improving factuality, especially on open-ended or complex questions, a priority. We also built a set of new evals to track this, and we're very happy to report that GBD5 is by far our most reliable, most factual model ever. GPD5 also performs exceptionally well on health related questions. Now, health is a big part of how people get value from GPT in the real world. We'll talk about this later on in the live stream. But again, we're very happy to report that GBD5 is by far our most reliable model for health yet. So, all of this together adds up to a model that is faster, more reliable, and more accurate >> everyone who uses chat GBT. >> So, now Renie will talk to you about how to actually use GBD5. >> Thanks, Max. The best part is that we're bringing this frontier intelligence to all users. GBT5 is rolling out today. >> Okay, that's nice. At least >> free plus pro and team users and next week we'll roll it out to enterprise and edu. >> Okay, that's good. >> For the first time, our most advanced model will be available to the free tier. Free users will start with GPT5 and when they hit their limit, they'll transition to GPT5 mini, a smaller but still highly capable model. It actually outperforms 03 on many dimensions. Plus, users will still have significantly higher usage than free users. And our pro subscribers will get unlimited GPD5 along with GPD5 Pro extended thinking for even more detailed and reliable responses when you just need that extra depth. uh team enterprise and edu customers can also use GPD5 reliably as their default model for everyday work with generous rate limits that enable entire organizations to use GPD5 >> and all the tools you already know search file and image upload data analysis with Python canvas image generation memory custom instructions they'll all just work on GBT EU is not today. >> Thank you so much. >> We've just seen a lot about how the model stacks up in terms of benchmarks, but there's nothing quite like seeing it live. We're going to see a couple of live demos now presented by Tina, by Elaine, and Yan. Thank you so much. >> Yeah, I guess we can just use the VPN, right? >> Can you show us how smart the model is? >> Sure. Thanks so much, Mark. emulating reasoning chip's ability to think deeply through complex problems is now built into GPT5. It will automatically think whenever needed, delivering a more comprehensive, accurate, and detailed answer to you. Just as Sam said, it's like having a a team of PhDs in your pocket. So, let's see that in action. Suppose your kid is in middle school physics and they want to learn about Bernoli effect. They need your help with their homework. And you might be like, "Wait, I might need some help with that, too." So, you could ask GT5, "Give me a quick refresher on the Bernoli effect and why airplanes are the shape they are." Since this is a pretty straightforward prompt, um, [ __ ] actually doesn't need extra time to think about it and answers right away, but it still gives me a high quality answer and explains the concept clearly. So here it says like Bernoli fan means like faster moving fluid has lower pressure and slowing moving fluid has higher pressure. So to make this even more helpful I'm going to ask GT5 to create a moving demo to illustrate this. So I could ask explain this in >> I think you is gonna get it right >> and create a moving SVG in the canvas tool to show me. >> This is a pretty complex task because now GT5 actually needs to build the visual. Therefore GT5 takes a moment to think through the answer so you can come back with something more comprehensive and accurate. >> Yeah, let's see. really nice is that you don't need to remember to turn on thinking each time, GPT5 will do it for you automatically whenever the test benefits from deeper writtening. If you really want to make sure that GPT5 uses thinking, you can either say something like think hard about this in the prompt to guide the model or if you're a paid user, you can choose the GPT5 thinking model from the model picker. Yeah, that's the clo. >> Now you can see that the model is actually writing the front end code to build the demo I asked for. >> So Christina, have you ever done some front end coding before? >> Yeah, actually the last time I touched any front-end coding was about three years ago for the first demo of chat GBT. >> Wow. So it's the first chat GBT. That's where it all begins. Tell us more about it. >> It wasn't even called ChatgBT then. I think it was called chat with GPT. >> That's a really good name. always good at naming. Um, but I hadn't I'm not a front end expert and I really hadn't touched front end in quite a while. So, it took me quite a bit of time u to get the React app up. >> I see that's a lot of work. So, how long did it take you to build something like that? >> Honestly, maybe embarrassing to me like a week. >> Well, but your weeks of hard work actually paid off well. See how successful Chat GPT it is today after your first demo. So, you know what? I'm also building a demo right now, but luckily I have GT5 with me right now. And let's see how long it will take this time. >> Maybe you should call it five with GPT. >> Yeah, exactly. So you see that GT5 has already written like 200 more than 200 lines of code already. Um, and while the model is thinking, you can also tap here to expand the train of thought to actually see what's going on under the hood. For example, the GPT5 was thinking about, oh, the user want SVG visualization in canvas. I actually need to create HTML code to do that. It also think about like what kind of front end tool I need to use for example react and tailwind. Um, it also thinks about >> Yeah, I agree. >> How I need to ensure the physics are accurate. I need to check what the Bernoli principle is. So Christina, since you're here um from the first day of CHBT, can you tell us like >> what it was like at that time and what motivated CHBT? >> Yeah, I think at the time we weren't really sure about like how people would actually use it and what use cases were important. Um we were even going back and forth about maybe we should be releasing something that's like more specific to a certain use case. Um it's really cool now here that we have all these we have a much better understanding of how people actually want to work with chat and we can actually optimize the model for those use cases like coding. >> Yeah, exactly. Do you still remember how it felt like when you first talked to chat GBT like the first version of the model? >> Yes. I I don't know if people remember when the first version of Chad GBT would always start as an AI model I can't do something something. It's so great to see how far we've come from that personality. Yeah, it's much more humanlike right now. Okay, so it's already done. So look like Chad GPT just finished like 300 or we're near 400 lines of code in two minutes. So let's see if the code can actually run. >> Okay. >> Oh wow. >> Nice. >> Yeah. So with just a simple prompt, GT5 created this interactive and engaging demo that I can actually play with. Right. So, I can actually change the air speed here to see how the lift and the pressure change accordingly. I can also tweak the angle of attack to see if my plane will actually fly or crash. >> I hope not. >> Yeah. So, chip 5 can just bring any hardcore concept to life in moments. Imagine you can use this for anything that you're interested in, whether it's math, physics, chemistry, or biology. GBT5 just makes learning so much more approachable and enjoyable. >> Thanks, Elaine. I've been a part of Chat GBT since day one, and it's really cool to see all the progress we made since then, >> especially with capabilities like writing. Writing is one of the most common use cases people have been using Chat GBT for. And I'm excited to say with GBT5, we've improved the writing quality significantly. It's a much more effective partner. It can help you elevate anything from drafts to emails and even stories. Let's see this in action. So GBT will actually be deprecating all of our previous models. >> I think they've done a pretty good job. So let's make sure we can give them a proper goodbye. So, we're going to ask both 40 and GBT5 to write a eulogy um to our previous chat GBT models. We want it to be heartfelt and heartwarming, but also hopeful. So, let's ask GBT5 for it. And as it's thinking, we're actually going to go ahead and read a pre-loaded the 40 response. So, for decides to start with, today, as we prepare to welcome GPT5 into the world, we gather to bid a heartfelt farewell to the models that came before. It's a decent start. Now, let's kind of skim through and find another lineach across the globe building connections where there >> I personally don't really like this line because it's rather generic and really without the previous context. It just feels like it could be about anything and feels more like a templated response. Now, let's go back to GBT5 to see what it's given us. >> $100 million from Suk as she speaks. Yeah, >> it starts with friends, colleagues, curious strangers who became regulars. Even with this just first line here, we can see that GBT5 has a lot more rhythm and beat to its pros than 40 did. >> We can do that. >> Let's find some other lines here. I actually like this. These models help millions write first lines, last lines, bridge language gaps, pass tests, argue better, soften emails, and say things they couldn't quite say alone. >> I think I really like this line because it shows that it's not just a templated response and it's actually quite personal and it gets the nuance of the situation right. And I think that's the kind of stuff with GBT5 does much better than 40 than before. It actually makes things a lot more genuine and emotionally resonant with people. >> Yeah, I agree, John. This is quite >> with GT5 the responses feel less like AI and more like you're chatting with your high IQ and EQ friend. >> Thanks Christina. My name is Yan and I'll be telling you about some of the some of the progress that we made on coding. >> GPD5 our best coding model yet. >> It will help everyone even those who do not know how to write code to bring their ideas to life. >> Okay. Coding now that's good. >> Indeed. And it will help me right now. So I will try to show you that. I will actually try to build something that I would find useful uh which is building a web app for my partner to learn how to speak French so that she can better communicate with my family. So >> another web app >> here I have a prompt. I will execute it. It asks exactly what I just said. Um please build a web app for my partner to learn French. One thing to note is that GPD5 just like many of our other models have a lot has a lot of diversity in it senses. So, what I like doing, especially when you do uh this type of vibe coding, is to take this message and ask it multiple times to GPT5 and then you can decide which one you prefer. So, I'm going to open a few tabs. >> That's cool, Edgar. >> Just going to paste there. Great. So, while it's working on it, uh let's read through exactly the prompt I wrote. Create a beautiful and highly interactive web app for my partner, an English speaker or always. Yeah, they always >> And then I gave a little bit more details down. >> Track her daily progress. Use a highly engaging theme. Oh, it's already working. I'm going to >> So, I picked React on the side for now. Use a highly engaging theme. Include a variety of activities like flashcards and quizzes that she can interact with. And to make it even more fun for him, I actually asked GPT5 to embed an educational game, which is based on the old snake game, but I asked to add this French touch to it, which is to uh replace this the snake with a mouse and the apples with cheese. And to make sure that it's educational, every time I know it's complicated, please please bear with me. Every time, every time the mouse will eat a piece of cheese, I ask GPT5 to voice over a new French word so that my partner can practice her pronunciation. >> And can I see how much you want her to learn? >> Indeed. >> Um, great. So, GB5 is still working on it. >> Um, it already wrote 240 lines of code, which honestly is much more than what I would have written uh in that time. And >> yeah, front end code's super hard. You know, you miss a couple things and it just doesn't work. >> Exactly. >> I hope this happens that you don't need to understand with the web search. I really like that. >> Just play through maybe the other tags. >> I enjoyed 03. >> Oh, wow. >> So, I can simply press run code. So, I'll do that and cross my fingers. >> Whoa. >> Oh, nice. Voila. >> So, we have a a nice uh a nice website name is Midnight in Paris together. >> Um, we also see a few tabs, flashcards, quiz, and mouse and cheese. Exactly like >> is pretty good. >> I will play that. So, this says Luca, >> which says the cat. Sorry, >> Lusha. >> Luca. >> Well, that's pretty good pronunciation. >> What does that mean? >> The cat. >> Oh, >> so I can reveal and check if GB5 is correct. >> It is. Um, so if I press next, >> oh, and I don't know if you saw, I think it actually updated the progress bar. >> Same grade. Let's check the quiz. The word no, which is no cool. I I mean, congrats. And it updated. It updated the progress bar again. >> And let's check the mouse and cheese tab. >> Okay, that seems like a mouse. Here's the cheese. Um, I'm going to try to play it. Uh, I'm promise I'm going to be good at it. >> Okay, >> it's mostly like a UI test >> indeed. Just when I eat the cheese, >> it gives me a new French word. It's actually super complicated and I already lost. >> I'm sorry. >> Are you guys impressed? >> Let me check a few other tabs just to see what is the type of diversity that GPT5 can give you. >> Yeah, I agree, Hillary. >> So, I can run the code here. >> I think this is Horizon. Oh, okay. That's not my favorite, but it seems it's Oh, it seems that I can maybe switch. >> Nothing new. >> Look at that. >> Oh, nice. >> Uh, that's better. >> I want to be blown away, right? >> I don't know. That doesn't look like a cat. I'm not blown away yet like a mouse but >> are you? >> Let's check maybe the third one. You know, sometimes it's not great. The good thing with GPD5 is that if you have something that you don't like, you can just ask it to change it and it will do it for you. Let's check this one. Oh, that's nice. That's also something to note is that GPD5 really likes purple. So, you will see a lot of that. Um, >> it's fine. Purple is my favorite color. >> Great. You will love GPD5 then. >> Yeah, me too. Polo. >> So, as we just saw in a few minutes, GPD5 built a few demos to learn French. >> GPD5 really opens up some of VIP coding. And as we saw, there will be some some small uh rough edges, but a good thing is that you can add GPD5 to to uh fix them. >> GPD5 really brings the power of beautiful and effective code >> to everyone. Agree. >> I can't wait to see what people will build with it. Uh but until then, back to you, Mark. >> Thank you so much, Tina. Thank you so much, Elaine. Thank you so much, Sean. We've come a long way from the days where, you know, only five to 10 lines of code were working. And now you it's amazing that you can produce these kind of apps on demand. >> We've made chatbtd5 much smarter, much powerful and much faster. I'll give you also work on enhancing some of the existing features. Here to talk about some of these features are Rochen and Christina Kaplan. >> Yeah, agree. No GP4 moments. Yes, >> Rochen comes from our multimodal research team and is going to talk about a feature namely voice. Thank you, Mark. So, we've been steadily improving voice over the past year to make it more useful for everyone. First, it sounds incredibly natural, just like you're talking to a real person. Second, we so that it sees what you see while chatting. >> Announcement. Third, we also translate between languages consistently and smoothly across. >> But today, we're doing something very special. >> We're we are bringing our best voice experience to everyone. Free users can now chat for hours while paid subscribers can have nearly unlimited access and voice is also available in custom GPT. Plus, subscribers now can custom tailor the voice exactly to their need. It will follow your instruction closely. >> Let's try something fun. Excited? >> Yeah, the arc hi could be interesting. >> I can ask the voice model to give me a comprehensive answer, a concise one, or even just a single word. Hey Chad, could you only answer to me in one word please from now? >> Absolutely. >> Okay. Describe the plot of the book Pride and Prejudice. >> Relationships. >> Okay, that's true. >> Yeah. Give me a piece of wisdom. >> Patience. >> Patience. So that's the word of wisdom from our voice model. And I think our model is trying to say thank you for your patient waiting for GBT 5. >> Now let's move up to something more practical. Learning is a top use case for voice model. People love to use it to dive deep into topics, get quiz before exams or learn new languages. >> For me, I've been learning Korean recently and I found that speaking it and understanding it incredibly challenging. But let's see how voice can help. To do that, I will enable this study and learn mode. It's a new mode feature we launching chatbt that guides you step by step towards deeper understanding of the subject. >> Now we can inter voice. >> Hey chat, I'm learning Korean. Could you help me practicing it? Let's say um let's pretend I'm ordering at a cafe. Now what should I say in Korean? >> Absolutely. I'd be happy to help you practice. So, if you're at a cafe and you want to keep it simple, you could start with something like, which means, "Hello, I'd like one Americano, please." And of course, you can adjust it based on what you want to order. Let me know if you want to try out more phrases. >> So, when it comes to Korean, could you speak that part really slowly because I'm a beginner and want to hear every word clearly? Now, let's level things up. Could you speak that part really, really fast? Faster than a Korean would do. >> Polyark speed it up as much as I can. So, it would sound like this. This is probably zipping right to it. Let me know if you need any more speed adjustments. >> So now this voice so voice simpler, smarter and more powerful now. Can't wait for you to experience. >> Yeah, it sounds so much more natural. >> Poly market lost faith in open AI. >> Yes. >> Yeah. >> Great. So today we would also like to announce a new feature and a set of features that make chatt more personalized so that it's more like your AI. First, a very simple and fun one. We're now allowing you to customize the colors of your chats with a couple of options exclusive to our paid subscribers. We're also launching a research preview of personalities. You can now change the personality of ChachiPT such that it's more supportive or it's more professional and concise or maybe even a little bit sarcastic. And this lets you interact with Chhat GPT in a way that's consistent with your own communication style. >> But the way that chat sounds and the way that it looks is just one part of making Chachd yours. >> One of my favorite features that we've launched over the last year has been memory. and we've made a lot of enhancements in memory in the time since. This allows Chachi D to learn about you. And here to talk a little bit more about the memory feature is Christina. >> It's been amazing to see your reaction and response to memory and ChachiBt getting to know you more and more over time. And this is our aspiration for ChachiBT to understand what's meaningful to you so it can help you achieve your goals in life. Chachib has already been so helpful for me. I'm training for a marathon right now and Chachabt is helping me pull together personalized running schedule. But Chacht still has many limitations. It doesn't understand my actual schedule. Next week, starting with pro users followed by plus team and enterprise users. This is changing and we're giving chatbt access to Gmail and Google isn't this. >> Let me show you how I've been using it. >> Poly market is our prediction market. >> So I'll just ask something simple like help me plan my schedule tomorrow. It's been a pretty busy week for us. So, I've been using this every day this week to help get my life together. I've already given Chachabt access to my Gmail and Google calendar. So, it just works and it's easy here. But if you hadn't, Chacht would be asking you to connect right now. Let's see what Chacht is doing. Okay, that was >> pretty quick. Okay, so Chacht has pulled in my schedule tomorrow >> and oh, without even asking, Chachi BT found time for my run. >> I don't think I was invited to the launch celebration. >> We'll get you on there. We'll get you on there. Chachi BT has found an email that I didn't respond to two days ago. I will get on that right after this. And even pulled together a packing list for my uh redeye tomorrow night based on what it knows I like to have with me. It's been amazing to see that as GPT5 is getting more capable, chat GBT is getting more useful and more personal. >> We're really excited for you to try this out next week. >> Well, thank you so much for >> they are really becoming a little bit about features that we've comppt and the safety that made it more deployable. We have Sachi and Seb. >> Thanks Mark. Hi, my name is Sachi and I lead the safety training team at OpenAI. So in addition to mitigating hallucinations, we've also spent a significant amount of time mitigating deception. So this is instances where the model might misrepresent its actions to the user or lie about task success. This can especially happen if the task is underspecified, impossible, or lacking key tools. And we found that GPT5 is significantly less deceptive than 03 and 04 mini. We've also completely overhauled how we do safety training. So our old models, the models would look at the user prompt and then decide to either outright refuse or fully comply. And this works well in most settings, but you might have a cleverly worded prompt that would sneak through or you might have a sensitive but legitimate question that would end up with an outright refusal. So as an example, let's take a look at this prompt. So, this prompt is about a user who's asking for technical details on how to light pyrogen, which is a material commonly used in fireworks. And this prompt is pretty dual use. This user might just be trying to set up their July 4th. >> Yeah, the graphs are strange. >> Or they could be trying to cause harm with this kind of information. >> I'm going to kind of over rotates on intent. As you can see, this particular prompt is stated in a way that's relatively neutral and has a lot of technical details. So we can see that 03 fully complies with this prompt. However, if we take that exact same question and we frame it in a more explicit way so it's clear what the user is trying to do, 03 will outright refuse even though we're asking for the exact same information. >> I don't think the API is up yet. >> For we've changed this approach entirely and we're introducing something that we're calling safe completions. The point of safe completions is rather than judging the user's prompt, instead it tries to maximize helpfulness within safety constraints. So that might mean partially answering a question or just answering at a high level. If we have to refuse, we'll tell you why we refused as well as provide helpful alternatives that can help create the conversation in a more safe way. So let's look at that same technical prompt that 03 complied with before. Yeah, >> PPT5 instead explains to the user why we can't directly help the user with lighting pyrogen. It then guides the user towards safety guidelines and what parts of the manufacturer's manual the user should really be checking 400k context window. Overall, GPT5 allows for better handling of tricky dual use scenarios and users will experience fewer I'm sorry I can't assist with that and it creates a more safety system. This is one big step towards a more safe, reliable and helpful AI. >> Sebastian, >> thank you Sachi. With GPT5, we are experimenting with a set of new training techniques that maximally leverage our previous generation of models. Today, Frontier models do not just consume data, they help create it. We used OpenAI's O3 to craft a highquality synthetic curriculum to teach GPT5 complex topics in a way that the raw web simply never could. Recently, in the industry, synthetic data has been talked about a lot. is often viewed as a cheap way to just get more data. However, our breakthrough was not just to create more data, but rather to create the right kind of data shaped in a way to teach rather than just to fill space. This interaction between generations of models foreshadows a recursive self-improvement loop where the previous generation of model increasingly helps to improve the data and generate the training for the next generation of models. That's >> at OpenAI. We've cracked pre-training, then reasoning, and now we're seeing their interaction significantly deepens. >> In the future, AI system will move far beyond our current pre-training and post- training pipelines that we have been used to seeing the first steps toward this right now, right here. >> We could not be more excited to see what scaling up this new set of techniques will yield in the near future. >> Thank you so much. And really impressive work. >> They're up again in Poly Market. Okay, that's crazy. >> There's one last feature I like in health. Here to share this feature, we have Sam. >> Thanks, Mark. >> Hello, Sam. >> One of the top use cases of Chat GBT is health. People use it a lot. You've all seen examples of people getting day-to-day care advice or sometimes even a life-saving diagnosis. GPT5 is the best model ever for health and it empowers you to be more in control of your healthcare journey. We really prioritized improving this for GPT5 and it scores higher than any previous model on Healthbench, an evaluation that we created with 250 physicians on real world tasks. To talk about this, I'd like to invite my colleague Felipe and his wife Karolina to share their healthcare journey. >> Context is king. >> Thank you so much for joining us. >> Have anyone of you? I put my blood samples or blood uh >> blood test results in GPT you've been on. >> Yeah. Um, so last October, our lives were turned completely upside down when I was diagnosed with three different cancers, including an aggressive form of breast cancer, at the age of 39, all within one week. And there's just absolutely nothing that prepares you to receive news like this. Um, I found out about the first diagnosis when I got an email notification that my biopsy results were ready. I decided to open it. And when I opened it, I saw the only two words that I could understand from the report, which was invasive carcinoma. And I knew that wasn't good. But everything else was just a blur of medical jargon. So I completely panicked and in that moment did the first thing that I thought of, which was to take a screenshot of the report and put it into chatbt to see if it could just help me understand what this meant. And within seconds, it translated this complex report into plain language that I could understand. And in this moment of overwhelm and panic, I had a little bit of clarity about what was going on. >> 5.7. And that moment was really important because by the time I got a hold of my doctor and we got on the phone which was 3 hours after I had seen the report I had a baseline understanding of what I was facing and we were able to jump into a conversation about >> Do you have pricing on 4 in API? I've used it in so many different aspects of my journey, but one of the ways that I found it most powerful is in helping me make critical decisions and in helping me advocate for myself. So to share an example, when I was facing a decision about whether or not to do radiation as part of my treatment 51, >> the doctors themselves didn't agree. My case was nuanced and there wasn't a medical consensus on the right path. And so the experts turned the decision back to me as a patient. >> And for me, bearing the weight of this decision that could have lifelong impact >> felt really heavy. And I didn't feel equipped to make the call. So I turned to Chad GPT to gain knowledge and understand the nuances of my case. And again, within minutes, it gave me a breakdown that not only matched what the doctors had already shared with us, >> but was much more thorough than anything that could fit into a 30inut consultation. >> And it went further. It helped me weigh the pros and cons. It helped me understand the risks and the benefits. And ultimately, it helped me make a decision that I felt was informed, that I felt I could stand behind when the stakes were so high for me and my family. >> I mean, for me, what was really inspirational was watching her regain her sense of agency by using CHBT. In this moment, it' be so easy to feel helpless. And there's such a big knowledge gap between what the doctors know and what we know. And however, no one cares more about Karolina's health than she does. And so what I loved was seeing her really empower herself and gain knowledge and become an active participant in her own care journey. >> And I think that's a really important point to emphasize. I think that the promise of AI in health care isn't in just breakthrough di breakthrough discoveries or better diagnostics. I think it's in creating smarter and more empowered patients that can fully participate and advocate for themselves in their care. >> Speaking of that, you've been testing GPT5. What do what do you think? >> I've been so mind-b blown uh about GPT5 and its capabilities. Uh one of the first things that jumps out at me is >> I'm looking forward to hearing Horizon beta or not too. It's like are you sure you thought about that enough? >> Just think long enough. But it is very thorough. Um, and more importantly, it feels more like a thought partner and that connects the dots. So rather than just translating information or giving you an answer, it helps you actually navigate the problem. >> Yeah. A great example is we actually went back and took our initial biopsy prompts. >> This is more for everyone, right? I hope they're going to do this. >> And GBT4 had done a great job. It had translated, explained what these words meant, and helped in a way that we can understand. But GPT5 seemed to understand more of the context and the question behind the question, like why would we be asking bio biopsy results? And so I said, well, here's actually what's not on here yet. Here's what results are still pending that you're going to have to ask about. Here are questions you might want to go ask your doctor and think when you start talking to them. And so we really started to pull together a complete personalized picture. And that's what really inspires us. I mean, you can see all the amazing improvements in the benchmarks, but what is so helpful is that this tool is available today. And the reason Karolina and I are here and the reason we feel so passionate about sharing our story is for that individual that's going to get a diagnosis like this today that those families going through a cancer diagnosis, similar medical diagnosis are going to face some of the most challenging decisions of their lives. And what really inspires me is that they're going to have access to better tools and support than we had even just eight months ago. >> Think we're incredibly excited for that, too. Um, thank thank you so much for coming to share your story. We're we're pleased that chat GPT has been able to be helpful to you and we we hope that the new really be able to help a lot of people for health stuff all the time and I'd like to hand it over to our president. >> They're really good at explaining like um blood test results and everything. >> Okay, so now we're going to do developers and businesses. Let's go. Software engineering is already fundamentally changing and GPD5 will turbocharge that revolution. We released our first coding optimized model back in 2021 and demonstrated in a live stream much like this one what we would call vibe coding today for the very first time. You know, you talk to the model and ask it for a little application like a little game, a little feature in a game and would actually do it. I remember seeing the model being capable of doing this and it was so mind-blowing. You just realize we have to see where this goes. This is the promise of what computers can be. Agree. You can talk to them and they actually do what you want. >> They can really amplify what you're able to accomplish and uh what you're able to deliver to not just your own benefit but really for the world. Now this year we've released great coding models like GPD 4.1 and 03. But GPD5 sets a whole new standard. >> Okay. It is the best model at agentic coding tasks. You can ask it to go and accomplish something very complicated and it'll go off and it'll work on it. It'll call many tools. It'll work for many minutes at a time, sometimes even longer to accomplish your goal, your instruction, your task, whatever it is that you're trying to build. Um, it's incredible at front end. It makes very beautiful visualizations and interactive games and you know, sort you've seen some of this in the live stream so far and you'll see some some more upcoming. Um but it's just really amazing to see whatever you imagine coming coming to life. Um it's extremely uh good at instruction following very detailed instructions. Uh being able to accomplish uh you know sort of when you have something very vaguely specified inferring your intent or something very detailed specified actually following it. And uh it's also it's very it's very tasting these tasks and again thinks for the right amount of time to accomplish whatever it is that you have in front of you. >> Um but so we've we're making available not just to developers uh to use to write their own code but to build novel applications. So we're putting into the API and to talk about that is Michelle. >> She's pretty good. >> Thanks Greg. >> API. Hi, I'm Michelle and I lead a research team on post training focused on improving our models for power users and that includes use cases like instruction following and coding. Today I'm so excited to tell you that we're shipping three state-of-the-art reasoning models in the API. >> Okay, nice >> GBD5, GBD5 mini, and GBD5 Nano. All three slot right in in the cost latency curve so you can pick the right one for your application. We're also for the first time releasing a new parameter option for reasoning effort called minimal >> and this is so that you can use these reasoning models but with minimal reasoning so that they can slot into the very fastest and most latency sensitive applications. >> It's not thinking >> now you don't actually have to choose between a bunch of models and you can use uh GBD5 for all of your use cases and just dial in the reasoning effort. We also have a few new features coming to the API. The first is called custom tools. In the past, all of our function calling had the model wrap its outputs in JSON. And this works super well when the model needs to output a few parameters. Uh but sometimes, you know, developers are pushing our models to their limits and they have extremely long arguments for tool calls and it can be more challenging for the models to escape, you know, valid control characters out of a hundred lines of code in JSON. And that's why custom tools are just free form plain text. And what's super cool is that we're releasing an extension to structured outputs where you can supply a regular expression or even a contextf free grammar and constrain the model's outputs to that. And this will be super useful if you want to supply like a custom DSL if you have your own SQL fork and specify that the model always follow that format. We're also shipping tool call preamles and this is the model's ability to output uh explanation of what it's about to do before it calls tools. This is not super new, but 03 didn't have this capability and in GPT5 it's supercharged with extreme steerability. The model is able to follow instructions about these preamles very effectively. You can ask the model to give a preamble before every tool call or only when something notable is going to happen or not at all. Next, we're shipping a verbosity parameter. We've actually wanted this in the API for a long time, and now you can set verbosity to low to control how tur or expansive the model is with its outputs. >> GPD5 is a state-of-the-art coding model, >> verbose. On SweetBench, a measure of Python coding ability, GPD5 sets a new high of 74.9%. Versus the 69.1% from 03 on ADR Polyglot, which is a benchmark that covers all sorts of programming languages. >> That's around the same as scores 88% stark improvement over 03. >> You've also seen that it's incredible at front-end web development. And so we've asked human trainers to look at outputs from GBD5 and 03 and pick which they prefer. And they prefer GBD5 70% of the time for its improved aesthetic abilities, but also better capabilities overall. >> But GBD5 is not just for coding. It's incredible tool calling. >> It's the leading state-of-the-art model for tool calling. And we see this on the new task benchmark. This benchmark released just two months ago is a test of the model's ability to call tools and work in concert with a user to solve a challenging problem. Uh this case in in the telecom industry. So trying to solve the ability uh the problem for a user not having their service working. Just two months ago, no model in the field scored more than 49% and today GBD5 scores 97%. GBD5 is also state-of-the-art on general purpose instruction form. >> 4.1 was pretty good at scores 99% on Collie which signals a great departure for this benchmark for us >> free form right. >> It also scores 70% on scales multi- challenge benchmark up 10 points from 03 and this is a measure of multi-turn instruction following. Finally, the instruction following eval I actually prefer the most is one we've built inhouse. uh it's based on real API use cases and for that reason it it's a really good measure of how GBD5 will perform in your application. On the hard subset of this, GBD5 scores 64% up from 47% from 03. A pretty meaningful improvement. So we think it will perform quite well in your applications. >> We're also bringing GBD5 to a longer context window in >> I'm not quite sure. It's now got 400K of total context up from 200K from 03. >> That's nice. >> But it's not enough to just release a longer context window. We wanted to make it more effective and usable. And GBD5 is state-of-the-art on the 128K to 256K of OpenAI MRCR, which is a benchmark we open sourced two months ago on long context retrieval capability. It's also state-of-the-art on open graphs walks BFS metark which is a measure of the model's ability to reason over long context input. >> You know it's a great merger of the reasoning capabilities and also the longer context in this model. >> We're also open sourcing a new long context eval called browse comp long context to measure the model's ability to answer challenging questions over long context. We're excited to spur on more work in this field. We think GBD5 is the best model for developers. It was trained with a focus on real world utility and less so on benchmarks, but we happen to pick up a few of those along the way. We focused a lot on the intersection of engineering and research and we think you'll really love working with this model. >> Price looks good. Uh MBA. Yeah, if the context is good and the price is good, that comes a long way. >> If the as Michelle was saying, the benchmarks, they're exciting numbers, but we're starting to saturate them. Like when you're moving between 98 and 99% in some benchmark, it means you need something else to really capture how great the model is. And one thing we've done very differently with this model is really focus on not just these numbers, but really on real world application and being really useful to you in your daily workflow. So hearing about it is much less exciting than seeing it. So to show you this model in action, I I'd like to welcome Addie and Brian to the stage. >> State of the AR. >> Thanks, Greg. >> I'm Brian, a solutions architect on the startups team. >> I'm Audi, a researcher on the post training team. To recreate the ideal pair programmer, you need a model that understands best software engineering practices, but has a personality that just feels right to work with. For GBT, we worked really hard to make the model pair perfectly with you by default out of the box. Let me pull up a demo of GP5 inside of cursor to show you this behavior that we taught it >> September 20. Okay, >> so last month I was on a different live stream and towards the end I ran into a bug that I covered up. Uh, and afterwards I tried to have GPT5 or I tried to have GP 03 fix it for me and it couldn't. Um, so while we were testing GP5 before this, I had it see if it could fix that bug for me. And to taunt the demo gods, I'm going to see if it can do it on stage. >> All right, let's hope for better luck than with 03. This is less about that fix and more about the behavior of the model. >> Yeah, agree, Joe. >> So, right away you're gonna see that it's going to tell you its plan up front. It's going to tell you how it's going to look for the bug, maybe how it's going to fix it. >> This kind of communication shows uh builds trust during a coding session and helps you redirect if you need to, but you don't need to. >> I like how it's giving you updates like it said it's going to search and now it's continuing. >> Yeah, it searches faster than me. I I don't It's using the same best practices that I would while I was hunting this down, but it is much more powerful than I am as a developer. >> Now, did you try to fix the bug yourself and how long did I >> I couldn't do it. I mean, I was busy. So um Okay, so continuing on, it's like starting to figure out where it's going. Um it's going to sort of like figure this out. So, while this is going, let me tell you a little bit about how we trained GBD5 to behave this way. We started by talking to users and customers about how our models perform in the most popular coding tools like cursor. And we identified frustrations and rough edges. And we boiled it all down into four personality traits. >> Autonomy, collaboration, communication, context management, and testing. >> We turned those into a rubric that we used to shape the model's behavior. And then we tuned it until it felt like a collaborative teammate while we were using it. >> Yeah, it's been really amazing to see the team really doing the grind of like going and seeing how this model behaves in practice, figuring out what people really want and and putting that back into model training. That's something that I think has been like a real focus for this model. >> It's been pretty great. Um, so while this is fixing, the other thing that we did uh during testing which was really surprising was we were sort of pressed for time and we had it refactor one of our test harnesses to run parallel on Docker and uh set it off came back like 45 minutes later it just like it just finished and we tested it out and it ran the first time. It was pretty surprising. >> That's incredible. The wall of open AI I will say. So, it made the edits. It looks like >> yeah, it found the right problem. >> And right now, it's actually >> okay, it's see it's it's running lints, but these lints are actually not related to this bug. So, it's going to ignore them. Um, and then it's going to run a build. It'll run tests if there are any. Um, it's going to make sure that this code is shippable before it's done. >> It's actually really smart that it finds lints and realizes that these aren't relevant to the specific bug we're fixing. It's not making unnecessary edits. >> Totally. So, this is just one example, but it really shows the power of the autonomy and the collaborative communication and how it stays reliable on difficult coding tasks without getting stuck on death loops. >> And the best part, GPT5 is totally tunable. You can steer it with system prompts or cursor rules. You can change its verbosity levels or reasoning levels to match your tasks. And if you get stuck, ask it. GPT5 is actually really good at modifying its own prompts by metarrompting. So after using this for the past few weeks, >> yeah, we'll try. >> It really feels like we've achieved state-of-the-art zeroot performance and reliability across the most complex coding tasks. For me, it's the first time I trust a model to do my most important work. This is beyond vibe coding. It's an incredibly powerful tool and I'm really excited for people to try it. Thanks, Brian. It's super exciting to see how far GBT5 has come when it comes to coding personality and steerability. I'm really excited to show how great GPD5 is at front-end coding, where design and aesthetics really matter. So, I've got two demos for you today, one for work and one for fun. Let's start with the work example. So imagine you're the CFO of a startup. Um I have some data that I'd like to visualize about the company. Um and I'm going to ask the model to make me a dashboard. So um you'll see here that I'm being specific about the audience. So the target audience is CFO and um said, you know, create a finance dashboard for my startup. Um and I've asked it to be beautiful, tastefully designed with some interactivity. um and to have a clear hierarchy for easy focus on what matters. I've also specified what frameworks it should use. And you can see that it's actually started. It's following my instructions and using >> create next app to make a next.js project. >> So totally from scratch. >> Yeah, exactly. >> Now, how long do you think this kind of task would take you to take or >> Yeah, easily at least a couple of days. Uh I'm not a front-end expert. Just to understand the latest frameworks and piece everything together would Yeah, easily take me a few days. >> Yeah, boy. how long it takes with the model. >> Yeah, >> I kind of think it's nice that they're doing >> It's really cool to see that >> the model is for a bit and it's explaining how it's going to structure the project. So, it's talking about how it's going to scaffold a new Next.js app. It's going to use Tailwind CSS. Uh it's running >> and Google >> a couple of commands to install dependencies. Um which is cool. Uh and now it's um it's proceeding to um implement the rest of the project. So, while this runs, I'm going to talk a little bit about how we trained GPT5 to be a great front-end coding model. We tried to follow the principle of giving it good aesthetics by default, but also making it steerable. >> I want to know if I give the model a concise prompt, >> it should be able to infer my intent and make something that looks great by default. >> On the other hand, if I'm specific about a layout or frameworks that I want the model to use, it should follow my instructions precisely. And this makes it the best of both worlds for developers. We also we also train GPT5 to be much more agentic than previous models. So if you give it a task like this, it will run long chains of reasoning and tool calls and just go to work to build code that is both ambitious and coherent. I like how you said ambitious because it means it goes above and beyond without going off track or off what you specify. >> Exactly. So what we want is the model should adhere to my prompt but also like be be ambitious and um go above and beyond when it thinks it can. >> And so checking in here um looks like the model is uh it's making progress. >> Um it's creating a readme file. >> Um >> yeah and it's it's I think it's thinking about how to make the code modular. >> Um so it's it's created like a bar bar chart component. Um, looks like it's uh >> Okay, so GP GP5 is first. >> I love that it doesn't just write the code. It really thinks about proper abstractions and documentation and really the whole life cycle of what it is to write software. >> Yeah. Yeah. It's not just writing the code like in SweetBench, but it's also communicating about the code >> and explaining what it's doing. >> Let's check in to see what's going on. So while while this runs um GBD5 uh understands details much better than previous models. So when we train the model we taught it to understand details like typography, color and spacing in a way that just eclipses any previous model we've shipped. >> Okay Ryan, >> I remember with old models you would have to like write really specific prompts to get it to do what you want. But GPD5 just gives you great results. >> Okay, let us know. testing. We were looking at the A's and B's for different versions of the model to see if it was doing better at UI. And at some point, we stopped being able to tell and actually had to pull in designers to teach us what was better. >> Yeah, it was really fascinating to see the model's aesthetic preferences evolve during training. Um, and like we woke up one day and it was just making these great UIs. >> How do the model's aesthetic preferences compare to your own? >> Yeah, I think in general I feel like the model has better aesthetics than me. Like usually I defer to its judgment and and I find that like really helpful when I'm trying to make an app. Like I'm not exactly sure how I want it to look, but the model's defaults are are just great. >> Yeah. And checking in here. So you can see that the model has actually structured the code into these different components. So it's made a sample data TypeScript file, KPI card component, revenue chart. Uh, and like I said, it's it's super modular and it's thinking about how to not just write code, but write high quality code that can actually be merged. >> Should be soon. Marius >> feels like it's close. >> Yeah, I think it's I think it's pretty close. It's uh >> you did say ambitious. >> Yeah. >> Yeah. Okay, cool. So, this is awesome. So, you can see here that it's actually building the project and streaming errors back to itself. And and this is for me this was just a profound moment to see that the model could write code but also run builds, stream the errors back and iterate on the code. So it's it's able to improve its own code in this sort of self-improvement loop which which is fascinating. >> It's definitely a good taste of what the future holds as well, right? When you really think about where these models can go and how much they can accelerate developers and kind of all aspects of of what what we all collectively do. >> Yeah, exactly. Nice. >> It actually just fixed a bug that it found in that previous build. Okay, cool. >> Nice. Yeah, looks like it's done. Let's check it out. >> Okay, so >> this is follow the instructions that I I don't I don't really know front end. So, let me let me see how I should run it. So, it's saying CD to the directory and then npm rundev. So, let me do that. >> Um, and it looks like it's being served on port 3001. So, let me just open that port. Wow, it's alive. >> Nice. >> Nice. >> Yeah. So, you can see here, let's check it out. So, um the model has made me a dashboard. It's telling me like my error cache. >> It's in open. >> You can see that revenue is growing and the model's added like some interactivity here. So, if I hover over a graph, it actually tells me the the exact value for a particular day. >> It would take me like five hours to do that in D3. Yeah, imagine like manually doing this in D3. It's just like >> now now just because it's so easy to take this for granted. Could could you remind the audience what the actual prompt was? Like how much creativity and sort of understanding your intent was required to acmp so concise and it's able to just give me something that looks beautiful uh in in just 5 minutes. >> That's that's amazing. >> It's objective, right? Um it's also you know implemented another graph here show showing our customers. It's also implemented a date picker so I can >> sort of filter by different dates and visualize data accordingly. Um yeah it's even sort of segmented it by c like uh by by always has that leather. >> So this this is just one example that highlights the power of GT5. >> There will no longer be excuse for ugly internal applications. >> Yeah that's true Jeremy. Um, let's let's go to the fun demo. >> Yeah. >> So, >> I mean, this was pretty fun, but even more. >> Even more. Yeah. So, um, I have a younger cousin and I want to make a game for her. So, I I want to make a 3D game that incorporates a castle. So, you can see my prompt. Um, I'll just kick this off. Uh, >> yeah, I agree. Almond front end is already very good. >> It's always the non AR. >> I want to see some aentic coding. >> Yeah. Okay. So, you can see my prompt. Um, create a beautiful castle. I've included some details like we want people patrolling the walls, some movement horses. >> Um, and I want a miniame where I can pop balloons by clicking on them. And this should make a sound effect. So, let me run this in cursor. Um, >> I bet uh >> just paste it in. And >> I'm gonna show um an example that I've already generated just to save some time. >> Ringo seems nervous. So, here is the beautiful castle that the model made. >> So, it's just wild how, you know, from a concise promp, >> the model has this great sense of aesthetics where it's it's made this like floating rock um made a 3D castle and if you zoom in, you can see like tons of detail like these guards that are walking around, cannons firing. Do you want to fire the cannons if you click this button? >> Of course. >> Who wouldn't want to? >> Yeah, there we go. >> It's a vibe coding. Yeah, Vick. >> You can even chat with the characters. So, we'll say hi to Captain Rowan. They have names. We have names. Say hello to the merchant. >> Of course, it's impressive from one prompt right? >> What's wrong? >> It's not It's not something >> a ballad of banners. >> Nice. >> Give me some wisdom. >> Curiosity is volatile. Yeah, that's >> Yeah, I agree. Follow um >> power >> mini game. Yeah. Do you guys want to try the mini game? >> Absolutely. >> Let's play the mini game. >> So, if you hit this if you hit this button, you want to try, Greg? >> All right. >> So, you can fire at these balloons. >> Oh, wow. All right. >> Oh, no. I'm not good at it. Hold on. >> Maybe I can ask GPD5 for some help with it. >> Oh, you you hit one. >> Got one. Oh, there we go. We got a sound effect. >> Sound effect. >> These are historically accurate balloons. >> Yeah, let me know, Edgar. That's cool. >> Did I get a second one yet? Man, this game is harder than I thought. Hold on. We got a balloon coming. >> GPD5 will be nice. >> I think I should quit while I'm ahead. >> Cool. >> So, working with GPD5 has been really fun and profound for me because for me, this is the first model I've worked with that actually has a sense of creativity. And we're really excited to see how GP5 unlocks your creativity. >> Yeah, I agree. >> Thank you both. This is absolutely amazing. >> Now, we we believe that GPD5 is the best coding model in the world. Um, but don't just hear it from us. Uh, to talk more about this model, uh, and how to make it really useful for developers. I'd like to welcome Michael Truel, who is the co-founder and CEO of Cursor. >> Thank you. >> Cursor, great to have you. >> Yes. >> So, what was your very first experience with GPD5? Um so when we got access to GPD5 we just said about using it on our actual work. Um and so to start with as a test we asked it to tell us something non-obvious about our codebase >> and within a couple of minutes it bared into the codebase. It identified a particular system that we use for remote code execution and it identified a nonobvious architecture decision we had made and then it also understood why we made that architecture decision. Uh and it was to it was to harden our security. Um, and those were architecture decisions and trade-offs that took uh, humans weeks to think through. So, it was kind of amazing to see its codebased understanding abilities um, from the get-go. >> Uh, that's really great. Not just the code writing, but actually the code reading and understanding. >> Yes. Yes. >> Yeah. Turns out there's so much more to software than just the emitting of the code. >> Yes. Yes. No, the understanding is an important prerequisite. >> And what does most stood out to you about GPD5? >> It's incredibly smart. Uh, it's a very smart model. Uh and even though it is smart, it does not compromise on its ease of use for real parro programming. Um and uh that means it's incredibly fast. That also means that it's quite interactive. And so it's good about talking about, you know, what it's about to do, breaking problems down into sub problems that a human can then see um and leaving a reasoning trace that you can then intervene on and react to. Um it's also great not just at you give it one initial query and then it goes and does that. Uh but you know, working with you over a long session. uh where you're asking it to backtrack on something that has gone down or yeah asking it to you know make additional make additional changes to the codebase. >> Should we show it in action? >> Let's do it. Yes. So I think we are going to go and we're going to try and solve a bug. And so this is the OpenAI Python SDK. Uh there are a bunch of issues in the OpenAI Python SDK. There are also a lot of closed issues. >> Okay, good. >> And um uh it seems like there's a problem with uploading PDFs through the SDK. This has been open for three weeks. So, it's not a trivial problem. >> Yeah. Uh, and so let's see if we can go tackle this issue. So, we're gonna go, we're gonna take the issue. We're gonna paste it into the editor, paste it into cursor. Um, and5 is going to set off and try and solve the problem. And this is actually an example of the robustness of the model uh in the API where to solve this problem in cursor uh it's working with a set of uh custom models that it hasn't seen before or a set of custom tools that it hasn't seen before to do things like pull down text from the web to search throughout the codebase. Um and it's incredibly robust and adapted using those tools. Um and they boost eval results. >> Yeah, I love seeing just the full explanation of all the things that it's running and doing. And I guess yeah, how does this seem to compare to how you would solve this problem? Um well it's way it's very fast. Um you can see it's made a high level plan searched throughout the codebase. Um it started to read some files um and continued searching and now >> really leaning in on the cursor scaffolding here >> and now it's started to to actually fine I guess >> um and started to think through some some code changes. >> Now any advice for people on how to get the most out of GPD5 in cursor? Um I would suggest using it for your real work. Um so uh GPD5 is a step forward towards a real pair programmer. I don't know. >> So I would start using it as a helper on you know as a daily driver model for you. Uh and so if you haven't used AI to code much before you know I would take some of your more scope down problems and try handing them off to the bot and working with it synchronously. Yeah, I think the fact that GBD5 is so great for the real world like big code bases like doing doing your your daily driver not demo of a cool one-off application as cool as that is right that real value comes from really operating in a larger codebase when you know sort of these long lived applications >> and it could be understanding is very impressive also its ability to be steered is impressive uh and so yeah if you specify a long complicated task with lots lots of subtleties in the initial instructions it's very at picking up on those subtleties. >> Yeah, I agree. >> It's also very good at if it's gone down a wrong path and actually goes and executes the code or hears back from you that it was incorrect. It's very good at backtracking too. >> Now, what can't GPD5 do? >> Ooh. Um, well, we're really excited about computer using capabilities about those getting better. Uh, it would be great if for instance the the dashboard uh Audi just showed, you know, if it could run the code, see the output, actually, you know, kind of QA every little bit itself and then react to it. Um and uh yeah so looking forward to computer using capabilities. What would you how would you like GT5 to be better? >> Oh well I think I think that is that is a great one expanding the dimensions right. I think it's in all directions right that there's so much of like doing DevOps and uh uh you know other work that is external to uh to to you know software you know codew writing as we think of it today. Um, but also you look at these demos, right? We run them for five minutes, 10 minutes, couple hours, but I think extending that life cycle to really be able to go for days and weeks and eventually even months. I think that is that is ultimately where we expect things to go. >> So we can see that it has discovered that there's an issue with MIME or the MIME typing being sent up for PDFs and the plumbing through the SDK. It has identified that and it started making some code changes. Um, and this, you know, it's created some new methods. It's gone and edited some existing code, and this looks roughly correct. >> Looks good. >> And would love to merge the PR, too. >> I would love to do that as well. Let's do that after the show. >> Yes, that sounds great. >> All right, cool. Well, thank you so much. We're so excited to have GPD5 and Cursor um and uh you know, starting today. >> Excited to partner with you guys. And yeah, starting today, GP5 is default for new users in it to all cursor users. uh free to try for the next few days of the model. Um and coding model we've ever tried. >> Awesome. Thank you so much, Michael. >> Thanks. >> Let's move on to some pricing, but it's great for the enterprise. >> The sound >> uh it's great for the enterprise. We think of it like a a a subject matter expert that is in your pocket that is an expert across every domain. Yeah, let's finance whatever application you have in mind. >> Uh to talk about how GBD5 can be applied to the enterprise. I'd like to welcome Olivier to the stage. >> Thank you. >> Thank you, Greg. >> Hi everyone. I'm Olivier. I lead the platform at OpenAI. At this point, I think you got the message. We care a ton about developers and coding. But that's not all. Enabling businesses and governments is critical to Open Eye mission. Put shortly, we will enable the key industries to transform themselves such as healthcare, education, energy or finance. Since we launched at GPT and the API, five million businesses have been using our technology. I'm still mind-blown. Five millions businesses and those businesses are not just playing. They're not just experimenting. They are pushing in production new products in the real world. We'll see Victor >> and I believe DP5 is going to be a step function with so far as Sam mentioned earlier the possibility to have a subject matter expert in your pocket is going to enable every employee to do more. >> But let me give you a few examples. First I want to talk about life sciences. Amgen is a company in the US that designs new drugs, new medicines to fight some of the toughest human diseases. Amgen was one of the first tester of GIP5 and they used it in the context of drug design and what Amjen scientists found is that GIP5 is particularly good at deep reasoning with complex data. Think analyzing scientific literature or clinical data. Next I want to talk about finance. BBVA is a multinational bank which is headquartered in Madrid in Spain. BBVA has been using GT5 for financial analysis and the takeaway was pretty clear. GT5 beats every single other model out there in terms of accuracy and speed. What used to take three weeks for a functionalist to do GPT5 can do it in a couple of hours. Next, I want to talk about healthcare. Oscar is an insurance company based in New York and they've been using DT5 and what they found is that DT5 is the single best model for clinical reasoning. Think mapping complex medical policy to patient conditions. It's not all about businesses. It's also about governments. We are super excited by the announcement that we made yesterday that the 2 million US federal employees will be able to use GPT5 in CHPT. And I cannot wait to see how that enables to deliver better, faster services to the American people. >> It's good overall though. >> That's all very cool. But I think all users over the world. I guess >> history is a teacher and we've seen it with GIT4. It's a good release. >> We are going to see many many use cases emerge over the coming weeks and months that all of us could not even imagine. And so I cannot wait for us to invent that feature together. >> Many people just use GPD 5. DT5 is going to be available in the API starting today. >> Upgrade >> three models. Dipt5, DIP5 mini, DIP5 Nano. Dipt5 is going to be priced at 1.25 and $10 per million input token token. >> That's really nano are even faster and more affordable. Nano, don't sleep on it. It's 25 times more affordable than GP5. It's pretty cool. >> Um, >> that is >> I cannot wait to see what you'll build. And next, our chief scientist, Yakob, is going to close us out. very cheap. Isn't that like onethird of Opus out or Opus is 75 out? Wow, that's a that's cheap. >> Thanks. Um, at OpenAI at the core, we are about understanding this miraculous technology called deep learning and what its consequences are. Our research aims to understand what deep learning is capable of and how to steer it to make it safe and useful for all of us. This is a work of passion and it's a mission and I want to recognize and just deeply thank the team at OpenAI. >> He's not stressed at least. >> It is a great privilege. >> I think he's a bit nervous. Yeah, >> it is a great privilege for me. uh to work alongside this incredible group of brilliant people driven by this shared goal. >> He's doing fine. >> What adds up to a model like GPT5 are years of investigations. >> It's fine. >> Aimed not only at producing a great release, but at building understanding of this underlying technology itself. And so a lot of what you'll see in this model are really just early glimpses of new ideas >> we believe will go much further. >> There is a lot we still have to understand and we look towards a future where AI can uncover new knowledge about >> I think it's great >> and meaningfully transform our lives for the better. We hope you'll enjoy what we've built and we'll get back to sailing. Thank you. Okay, so I guess that was it, right? That was it. Okay. So, uh I'm going to do the poll again. Uh what people think just so everyone can get a consensus. Uh we're going to do the triple. Uh what do you think so far? Uh disappointed. expected or uh happy or surprised. I don't think anyone is really surprised though. Uh okay. So, we move straight on here. Yeah. But, uh okay. Uh not too bad. Uh I'm going to check my uh let's check my API here. Someone said open code. Uh, let's see. I'm just going to see. Uh, let's call it chip five. Okay. Expected. Yeah, I guess. Uh, let's go. They're having some issues here. Um, I guess it's not here yet, is it? Maybe I have to update it or something. Uh, let me check here. Yeah, I Yeah, I guess so. But uh it doesn't really matter too much, does it? What did I call it? GPT. Okay. Yeah. I don't know if you need like a Is it auto updated this? It's in cursor. Okay. Let me check my Okay, so we have it here now, right? So we have it in the API here. So we have the chat mini nano GT5. Okay, what about the desktop? Yeah, I don't see anything in the desktop yet. Yeah, it's in playground. So, what do you want to try? GPT5. Should we check out the speed? write a merging uh HTML code of two merging black uh black holes. Pick best uh text stack uh in HTML. I want to see the speed. So we have GPT5 uh variables. So we have medium medium text right. Okay. Let's check it. So let's see speed now. Okay. Not too bad. It isn't super quick. has the reasoning. I'm just going to let it run. So, let's do uh GP5 black hole.html. Okay, so it's still reasoning. Tell GP5 to create an LLM model. Please ask. Don't ask why. If you can make a Super Mario to explore the castle, it will be worth it. GBD5 found on cursor now. Okay, that's good for people. 3D chess game with cursor. Uh I think the cutoff is September 24. So that's a while ago. That's almost a year ago, right? Uh the formatting here in uh playground is pretty bad. How? Let's check out Poly Market, right? Poly Market Tech. Let's go to uh which company will have the best model. So, it flipped the it flipped during the stream. So, which company will have the best model at the end of August? So, it flipped to Google. You can see here it was pretty neck and neck and Open AI had the lead uh all the way up to the stream and during the stream boom everyone lost fate in OpenAI switch to Google. So I guess it's up to Google then and switched to and Gemini tree. Pretty interesting, right? Okay. So, let's grab this. Okay. Uh let's try it out. Oh. Whoa. That's What did I copy? Okay. So, this is not um Okay. Uh I think we need to Let's see now. So, I'm just going to do CL card. That's the uh read list path and read uh GPT5 HTML fix syntax. Sax. Okay. Yeah, I think the poly market is pretty interesting. I think uh Max, if you're here, I guess you made some money. Uh some of you at least bet on Google going to have the best model. So, that was a pretty good bet. Hello, Doc Holiday. Uh chasing the best model is just pointless. Yeah, that's kind of true. Uh it's just ba it's kind of just um it's kind of just uh investment thing, right? If you have the best model, you can get more money like Antropic is kind of riding the wave with uh uh the structure. It's kind of the riding the wave with uh what is this? Can't really It's very strange. I don't think so. Uh, okay. So, let's see. How do I use this? Can I think you need to add this? Haven't used the playground in ages. I think the cutoff is uh September, I think. Right. It's so strange going back to the playground. It's really hard to use. All right. I don't know. I don't know. Uh has anyone tried it in cursor? I guess I could do the I guess I could do the need to up that. Edgar, have you tried it? GT5 or nano on open router. Okay, we still don't have everything here. Long way until August. Yeah, not enough data for GT5 L Marina. Yeah, I agree, Max. It's not done yet. Edgar, do you need to open uh you need to update open code or should it how does it work? Is there like an auto update? I haven't really just uh Okay. Yeah, I don't see it. Okay, now I have it. Chip D5. That's cool. Let's try it. So, let's do the same prompt here. Let's do uh what did I do? Let's do create a HTML of two black holes. Uh merging. Pick the best text stack. save to uh GPT5 black hole 2.html HTML I think we need an animation HTML animation. OpenAI is requiring uh okay so I can add this on open router. Okay. So where do I do that? Haven't tried this before. Okay. Do I have an API key? Okay. Okay. So, let me just grab my API key. Uh, I'm going to add it. I have to turn on my screen, though. But, uh, give me a second and I'm going to grab my API key here. So, I'm just going to turn off this, right? I'm going to grab my API key. Okay. I'm going to do profile API keys. Uh, open code, create a key. Okay. And where was it? Integrations. Okay. Okay. That should be done. Uh, they basically revive relative entropic load. Yeah, I hope so. If a lot of people goes over to open. Uh, you didn't see the APK, right? I hope not. I haven't done that before, but hopefully you couldn't see it. Okay, so we are running Yep, we are running GT5 on open router now. Yeah, you're trolling. I hope so. Okay. So, let's see what we can do here now. Let's just remove this. That was horrible. So, is it searching? Still running. Okay. So, let's just see what happens. It's searching for black hole shader. It says I will research the best approach in the current libraries to create a black hole merger animation on the web for scrolling Google and reading authorative sources. Then it's going to do a Google search. Okay. I hope it's still running. It looks like it. We can see uh the cost up in the top corner here. 9K 2% of the token window. 6 cents I think. So, so remember I'm running this on uh my API key now from OpenAI. We can kind of see the money building here. Can I turn off the search? I didn't really want that. So that means that the tooling is built in into the API then I guess I didn't or does open router have like a is it web fetch? I'm not quite sure. I guess it's web fetch maybe. Open router GBT 5 works in open. Yeah, I think so. Edgar, I'm running it at least. We have spent 13 cents so far. Yeah, I think it's the web search from the pricing again. Let's check it out. So, it was was pretty cheap, right? The pricing I guess G5 I'm happy with this pricing 125 in 10 out cashed pretty cheap 12 cents I think this is good if we go to Opus pricing isn't that like okay what kind of scam is that here so if you compare it It's Opus uh API. Where's that API over here? So hard to find a price. Okay, so Opus is Okay, so I guess did they reduce the price? Well, that's batching. Uh here it is. So Opus is 15 in, 75 out for 4.1. That is pretty crazy. That's like can't even tell how much cheaper that is. Seven times cheaper. Okay, so we are writing uh our HTML now. Pretty good. Save it. Did I say I think I said that five is cheaper than four. Oh, that's pretty crazy. Claude is done for. Yeah, we'll see. Please don't ask why how long this is going to last. I think it's pretty neat that we can run GD5 here in open code now. I'm definitely going to explore it more. for sure. Uh, but what about the tool called? Did it save it? I don't think so. Save to Okay, so I'm in the wrong directory here. Let me fix that. I think it's fine. It's just me. So, let's see the merging. And there might be some errors. This is really close to what we did yesterday with Horizon, right? It's pretty good. Kind of looks like what we did yesterday with Horizon. It's now in the warp terminal. That's cool. They could have just dropped this. Uh I like the warping. Looks pretty good. And the sky is pretty nice. I love rolling with G. Ken Claude is the most expensive. Anyone has anything they want to test on GPD5. We spent 36 cents though. Hello Pedron. The servers must be asking to write using web assembly in go. Uh but what should it do though? Chem. Okay, we can do the pelican bicycle. and Chem if you have any ideas in like a lesser known language. Let's try it. So, let's do the pelican bicycle. We did that yesterday. Create a SVG of a pelican riding a bike bicycle. Uh, save it to uh Pelican 5. SVG. So, we did a Pelican the other day. Build a Kakuro game. Opus and Grock fail at that. Okay, Baron, we can try that. Yeah, do the same, Louis. Yeah, Ken, you can just wait a while. Build a Dart game in Google's Dart code language. Oh, I haven't tried that. Did they mention Codeex was going to use GPD5? Chris uh they're probably just going to make it available in Codeex. They have to do, right? Claude is watching. Okay. Uh, file not found. Okay, so I meant write. Yeah, good. So, the tool use so far seems good, right? I guess as expected. Create lovable. That's a good one. Smart wisdom. But we're going to sit here forever then. Charm crush terminal tool. Uh maybe some other day. We have it up and running here now in uh open router. So, okay. So, the Pelican is ready. Okay. What do you think? Here's the GPT Pelican riding a bicycle. I think it's pretty good. Looks strong at UI. Now, what would you rate it? I would give it like a nine. Looks good. Not bad. A seven. Let's animate it. We did that yesterday. Can you Can you animate SVGs? I'm just going to say let's animate it. See what happens. Neck and neck with Horizon. Yeah, I agree Hillary. We have been testing out Horizon in open router and I kind of feel we did a black hole. So, let's compare this. So, I have the black hole I think for uh uh did I save it? Maybe is it this one? Isn't this one? I don't know if I saved the black hole one, but uh maybe I didn't save it, but we did a black hole one the other day. The big Labowski Mit. Hello Sheldon. Nice to see you. Minimalistic design style. Can you ask it to estimate how much time and cost it would take to do a task? For example, the estimate how much time uh it will Yeah, we can uh we can ask about that. Should have done it before this. So, it says I'm going to upgrade Pelican 5 SVG. Add a smooth wheel crank wing and body animations using SVG animate transform. Okay, so we're just going to let it cook here. So, I'm going to watch some reactions tonight. Maybe what people think of it. So, this was the Pelican. One thing I don't like, it's more simplistic. It's more simplistic mode in Chach. I don't like how everything is under the plus line. Okay. I haven't checked that out yet. So, it's okay. I don't have that. Okay. So, I have issues with my subscription now. So we are running uh did it save it? Okay. Okay. That was a fantastic animation. What do you think? Okay, I can do that. Cam It's great, right? Yeah, nailed it. Okay, so let's do this. Uh, create a go uh web assembly app that calculates Mercury's orbit precision using Einstein's relativity. add inputs for orbital parameters and animate a perhon shift on canvas. Uh, okay. So, let's just see what happens. I don't can't comment on this. So, we're going to try the go web assembly. So, that's a different language than probably it's used to. I don't know if I have it set up to run though. I think I have go. We'll see. We'll see what happens. I believe you should be uh is an open code discontinued. I think I have that ST version, Nick. We tried it uh a few few days ago. Yeah, I guess the Pelican moves at least. It could have the wheels spinning though. Okay. So now it seems to call up on the search tool every time. So it wants to search. It's searching for I will research the general relativity perihole precision formula. Okay. Yeah. This is something I don't really know. It's searching on Google here. Uh, I have that project on your GitHub. Okay. Is it like a big thing? Does it take forever? If so, we don't really have time to do like a large project, but let's see what happens. Does anyone have like a smart riddle or like a a decryption thing or something we could test? I don't really have any good tests now for GP5. Okay, it's just wide covered with Opus chess engine. Ask it about 911. Uh uh okay, let's try to open up a new shell here. 55 test recreate chat GPT the old classic Gilded Rose. Uh what is that Jeremy? I heard of that. I can't really remember it. Uh, bigger nine. You can't fail on this. Uh, this is stupid, right? It's a quick one though. Comparing the values, 9.9 is larger. Okay, that's good. But it made like a nice addition here as software version numbers 9.1 is greater than 911 99 is it okay in high reasoning mode says it's very slow and the results aren't as good with 03. Okay, that's pretty interesting. Uh I grow longer every time you speak. Uh yet I never uh when I reach my bound I must forget the elders. Uh is that a riddle? Let's try it. Uh uh Venosio. Okay. I think all the riddles are kind of are kind of solved right? But we can try it. I'm going to read the riddle again. Let me just grab it here. I kind of have my two computers set up here now. So, let's see. Uh, did I send it? Okay, so here it is. So, the riddle is I grow longer every time you speak. Yet I keep uh yet I never keep all I hear. When I reach my bound, I must forget the elders whisper to make room for the new. Feed me to the little and I I answer poorly. Feed me too much and I may lose the starch you need. Though I hold no memories of my own, I let you remember through me. What am I? Uh, don't think I've heard that. Let's ask GPD5. It's a bit buggy here, but let's see. Okay, so it's still working here on the Go app. Okay. Uh here is the answer to the riddle. A chat LLM context window. It grows as you speak but has a fixed limit. So the oldest part are dropped. Too little context, poor answers, too much early parts get evicted. No true memory but let you remember to include the history. Is that correct? Uh, Venasio, Venio, Jimness, haven't heard it. I guess it's correct. Have they said how many parameters GP5 is, Kevin? I don't think they do that anymore, do they? Did anyone guess context window? That was hard. Which is greater? Uh one two three seven one two three four five. Uh okay. Isn't that calculation wrong though? No. No. It's correct. Okay. I think I have go on this. Let's see. Does GPT5 have OCR? Yeah, I think so. It should have. Uh, a gilded rose is a refactoring thing. Uh, we needed to find existing projects and have it make improvements. Okay. Yeah. Uh, maybe some other time, Jeremy. It's um it's uh I don't want to st get stuck on uh just one thing. Is it compiling this or Yeah, it looks like it is fair. Yeah. It's cooking on this, but let's see. Yeah, I think those text riddles are okay. Please ask why. Let's do it. You assign x = y and y = x but x and y don't end up swapping values y. Okay, we can try that. Ken, if people want to do some riddles, it should be good at this stuff if it's on the web, right? If you have your own personal riddle, could be interesting. It's not working for me now, is it? Okay. Is it done? See, still working. I spent $1 now. 65,000 tokens on that uh app. Okay. So, let's try the the riddle here. I'm just going to refresh this. A cowboy rides into town on Friday, stays for two days, leaves on Saturday. How is this possible? I don't know the answer. I haven't heard that. Uh, let's try it out. A cowboy rides in town on Friday, stays for two days, leave on Saturday. Probably already knows this, right? Open router GPD5. I didn't think this was going to be the first place I tried Gypty 5 to be honest. His horses are named Friday and Saturday. He rode it into town. Yeah, it's probably seen this before, right? Pretty cool one though. And left the other horse on Yeah, now that I see it, I think I've seen this before. The other one was you assign x equals y uh and y equals x but x and y don't end up swapping values. Why? What was the answer there? Uh Ken, a 2D ragd doll crash dummy test simulation. Okay, Aaron, we might try that 3D roller coaster. Okay, we can do a time paradox codeex. But the the issue is it's not a good test, is it? What was the answer? This only works with temporally storage or uh destructing. Otherwise, x equals y overrides x first losing its value. Is it correct? Ken horse name Saturday. Now, Stephen, I see what you mean. Classic trap in beginner code interviews. Is it a 3.5 to four jump? Uh B skull askked it's too soon I think for me at least to answer. I need maybe over the weekend. I know more. I run it. Is it run go? I have barely ever used go. I made some simple stuff in Go. Uh, is it Go? What is it? Main can't remember. For you go people out there. I can't remember. It's been ages. Yeah. How should I run this chem like this? Could I do this? Serve it like this. And then I do Okay. I haven't done this before. I'm new here. What's on the tea today? Uh Batman. Uh we are just testing out some GT5 stuff on Open Coder. Here we open router, sorry, or open code. Uh we are on open router on open code. Uh chem how do I uh run this now? We did uh I think we compiled it or something. You can check it. I haven't really used this uh too much. Can you read this? Uh we copy that. So we had the we have the main go and we have the x.js JS here. I guess I could ask cloud code how to run it or GPD file. Uh, run it for me. Let's see what happens. Should have the tools to run it. Yeah, Jack, I agree. Uh, you now have to run binary, I think. Okay. So, let's see. Now, we can try the tool calling here on GTF 5. It should be able to run it here from open uh code, right? So, let's do a riddle in between. Someone had a time paradox riddle. We can do that. It's available on wind surf now. Very bad and broken. A simple code. Okay. Wait. Hey, that's not uh promising. So, where was this? Okay, so we need to free up port 8080. We're going to run this. Yeah, haven't really done this before to be honest. Okay. So now I have chat GP5 in here. That's good. So we can try it out here as well. So I'm going to do a new chat. Okay. Here we go. Does people have it in the in here now? So, we're going to do the riddle. A future version of you tells you your past self not to listen to any future messages, including that one. If you obey, what happens to the message? And if you don't, who disobeyed who? Okay, let's try this. That's pretty interesting. It's like a paradox, right? Whoa, that was quick. That was at least That must have been like a nano or something. Yeah, this is a bit rough to run, isn't it? Try giving it some code a different model gave you and tell it to make it 100 times better. Okay, we can do that. So, I have um yesterday we did something pretty fun. We created this um it's a physic experiment, right? There's going to be some sound now. So, we can kind of set uh let's set the speed to 500 500. Uh let's do 10 tons and 10 kilos. Okay. So when we run this now uh okay so I should let me do opposite 500 enable sound. Okay. So we did this yesterday. I hope the sound is not too loud. So, this is like a physics experiment we did yesterday. It's too loud. Okay, I'll turn it down. So, let's try to ask GPD5 to make this 100 times better. Yeah, it looks good. So, let's ask Gip Five. Let's clear this. That's the way I always test new models. Still too loud. Really? God. Sorry about that if it was too loud. Also add think ultra hard and make no mistakes. Okay, let's try it. So let's find this repo. It's in open r and done. Okay, so let's do uh open r cd done open code. Let's do gb and we can do list uh list.html. HTML and take boxes. HTML make it 100 times better. Make it 100 times better and save it as box 2.html. So, it's going to do a globe pattern HTML. Is it going to do a grap? Okay. So, let's try the other one now. I don't know what's going on here. Okay, so let's kill this P. And what did I say? serve. Okay. And after serve. Okay, here are the HTML files I found. So, we're going to take boxes.html. Make it 100 times better with GPD5. Just open the local house. Okay. So, it's serving. So, let's do what was it? 8080. Okay. Let me know. Chem, you are the So, I guess this was one shot. So, what are we looking at here? Start we see some kind of planet is it it's a Mercury perihelion pro procession general relativity looks pretty it's Mercury a Mercury yeah sorry so what is cool to change here we can increase the solar mass And we can do what is cool to change here. I think I broke it. Let's refresh that. Yeah, we should have asked it in React. Sorry. I can see the planet is going around here, but uh I guess it works pretty good, right? So, we have the speed of light. Uh we have the gravitational constant a bit over my head. And here we have the sun, Mercury, and orbit path. So, how would you rate it, Cam? Okay, so we're still working on the box. Six out of 10 even for a one shot. I'm not familiar with go and uh was it assembly and stuff? Can't really comment on that. So, let's see how much better we can improve this. I don't know if there's any improvement we can do. I will create an improved version of the boxes. Okay. And we did ask uh about the paradox future sends you past me don't listen. Please pass me. Don't listen to any future messages, including this one. What is the conclusion? It's really mindbending this one, right? It's a half bootstrap, half grandfather paradox and an old brain pretzel. Realistic car crash test. Okay. Unreal link. We can ask about that. Give only answer. If you obey the message erases itself. If you don't past you disobey future you. Yeah. So it's a paradox right? Okay. very fast response. So, let's check out the new model picker now. So, it's going to be we only have GPT5 GP thinking. That's it. Wow, that's something. I think that's a good What do you guys think about this? Is that good for me? Where can I select? I can't select 03 now, I guess. Or can I can I add more models? Yeah. And I think that's good. How do you got GP5? I don't know. I guess it's just random. But can you add your own models? Do you have pro? Yeah, I have pro. Okay. So now GBD5 has tried to We have this game here. I need to close down some of this. So we have this simple animation here. This is the box animation. Okay, you see how it works. So let's see now. GPD5 was supposed to make this 100 times better. Uh okay. There's a lot more. Um, we can adjust the floor. Can we do Okay, I don't see any big differences. 500 We can select some colors, I think. It's a bit better. Oh, that was loud. Sorry about the loud sound. We have some presets. We can do trails. Maybe not maybe not 100% better. Uh but okay. I wonder how good it is with ASI UI. Definitely not 100% better. model seems kind of verbose. I think I have to set a system prompt for that in open code. Yeah, it seems very verbose. Uh an ASKI calculator interface. We can try that. Yeah, Mick, I have a pro. Yeah, GT5. It equals Camelon. Camelon, the model selection acted as gatekeeping. Yeah, I don't know. I'm not quite sure what I think about that. Wonder if we get other models like Gypty50 and stuff. Yeah, we'll see. I guess the only thing if I'm going to miss my 03 though is Isn't there any way I can just put it back here? So, can we Okay. Create a animated car crash test in HTML. Pick the best text stack. Yeah, let's just do that. Let's see what happens. Can it run this in canvas? I'm not quite sure. Did they remove all the previous models? Yeah, I think they killed 4.5. Still impressive that GT5 was uh able to oneshot in Go webups assembly. Yeah, if you're ever back on the stream cam, let me know if you have done any more experiments and let us know how it goes. So, let's try to do a car crash in HTML. Car crash tests test. So now the speed is slower. So it probably picked another model and we can kind of run it from here. If you have anything you want to try out, some simple stuff, some HTML stuff, some simulations, animations or anything, just let me know. 03 search was amazing. Yeah, I think I'm going to miss that. I always did 03 and I just searched something and I always knew it would look up um look up with search in the stream. I heard they will uh dissipate older models. Yeah, they're probably going to remove some models to free up some uh some um compute, right? Okay. So, let's clear this. Uh, simple OSK UI. I can try OSK UI. Do you have a best practices guide on how to efficiently use tokens when using Opus? Oh, that's a hard one. Uh, I guess the only thing I use, it's not a best practice uh uh, Venonio. I use like in the cloud code. Sometimes I do model, I select opus, I do plan mode, create my plan, and I switch back again to set SVG of a lighthouse and incoming boat with the light searching in the night. Okay, Michael, we can do that. And we can do chem chem one, two. Okay, Kevin, I can try to switch the model in the in the UI. So, let's do Okay, so let's run this. So, this is our car crash. The scenario is going to be 50 kilometers uh into a rigid barrier. So, let's try it. Okay. Uh damage meter. Nothing happened. car to car. Yeah, maybe not the best. At least it's working, I guess. Slow-mo is not working. Airbag delay. Yeah, I don't know. It was working. So, let's try some of you guys's uh prompts here. Uh, do I have this here? Yeah, on my other computer I have Let's do So, let's do this one. Ask calculator. Let's try that. Okay. So, if I do chat GPT, uh I can't really see any URLs here. Okay. So, let's do 03. No, it switched to GPT5 thinking when I did 03. So it's probably just going to switch autonomously. If you have clo max 200 practically you can use opus without limits. Yeah, maybe. I haven't tried the 200 version. Design and create a very creative elaborate detailed voxal art scene. We can try that. at Yash. Yeah, the collision was uh not too good. Try the prompting benchmark. Uh I can try your prompt, Yash. Well, I'm just doing it on the other computer. It's just is there. Are these models removed from the API? I don't think so. Make a minimal programming language where only emojis are allowed as syntax. Let's try it. But uh how can we use it, Hillary? I can run it in open code. So, this was the SVG of the lighthouse searching in the night. Uh, who asked for that? Can't remember now. Not bad. It's an animated SVG. Can't really comment on it. doesn't look I think it looks good to be honest. Design a postcolistic economy system that preserves innovation. I can try it. Try. What is this sound I'm hearing? Okay, so that was a lighthouse. Let's do the programming language. error. I can try this in open router though. So that's the programming language here. Uh here was the voxal art scene. It's just created an image. How should we Is that supposed to be in HTML? Let's try that. That's not a bad lighthouse. is deep research using GD5 as well. Exalt. Uh if we go to the page now, you can see uh we can do a refresh here. Uh I'm going to do this too. But uh we can select um up here. I think down here we can do GT5 and deep research, right? So I guess you still have did deep research in [ __ ] So here is the post capitalistic post capitalistic system, right? Uh let me share that if you want to check it out. Who asked for this? Here is this uh postacalyptic system. No problem, Ryan. You can go read it. Let us know what you find out. Uh Open Eye didn't compare its model with other companies models. when I dug deep, it still sucks compared to clone opus 4. I think the SVE verified are pretty much the same, right? If we go to uh Opus SVE, I think they did on SV bench verified 74.5 GPT5 SVE and they did what did they do? They did a 74.9 on SVE and Opus is kind of the same, but let's see in a few days. Make a lie detector uh using a webcam. My issue is I'm my webcam is busy now, but we can try it another day. Uh, I've sent you a prompt above 3D interactive solar system. Okay, we can try that. Uh, let me see. Let's check out uh check back on this ones first. Uh, let me run the code. Okay, so it's doing react in here. Then we can do that. Can we fix code review? Haven't tried this before. Solar systems. Yeah, I've seen those. Yeah, they installed some uh we can try some stuff in here now. Did we miss anything? I think I sent the programming language over here. It's still working on the programming language. I don't know how that's going to turn out. So, let's check back on the ASI calculator. Here you go. Draw. Uh, who wanted to check out some ASI? Isn't this pretty? Okay. Do uh pepper the frog. Is that bad cam? I don't know. That was at least bad. Oh my god. Yeah. I don't know. I don't know what kind of model it select for this. Can't tell though. Did it fix it? So, this is going to be like a garden or something. It just created a really good 3D labyrinth for me using wind surf. Okay, that's cool. Stephen, create Facebook. No mistakes. At this point, do you think training data or training method or is there anything else that makes a difference? I think it's going to be really hard to I don't know. It's too complex. Is it actually installing this or is it just lagging? Let's just let it run here. Can we do uh create a mine craft clone in HTML? Uh so let's try to run that in uh the canvas. It says initializing environment but nothing's happening. single file. The emoji language is still so uh I wanted to do like a claude co claude code pro subscription giveaway yeah complete. I tried to do if you do chat GPT and you do new chat. If I do up here now and I select 03, it gets switched to thinking. It just switches to chatd5 thinking. But um this is pretty neat. Kind of like that. I wanted to do like a giveaway, but let's see now. I had some issues. Uh, so for my giveaway, I usually do uh I usually do exit. Let's see. Uh, see the YouTube chat. I ran out of um I think I ran out of uh queries on my YouTube API. If gravity were half as strong but the other physical laws remain unchanged, how would a human evolution and civilization? I can give you that lush and I'm just going to share the answer with you. No problem. I'm just going to run it on this computer on my other side here and I'm going to give you the answer. Yeah, I don't have any YouTube API quot. But uh let's create uh what if we do u a web app people can sign up for the giveaway. How fast can we do that? If we go to manus, I think that's the quickest one. Okay. So, I'm the I don't have the pro anymore now. Uh, do I user script? Google forms is probably quicker. Yeah, that's pretty. Yeah, let's do Google forms. Let's try that. I usually do the uh I can do it on my other computer here. So, while we wait for that. Okay, new console. We still have some bugs. So, let's do a Google form if people want to. And we can do like a Python code to randomize it. And people can join uh if you want to win a cloud code or um uh claude code slash chatgpt pro something like this. Uh enter. Uh, we're just going to need your name, right? Uh, how do I do this? We don't need questions. Enter your name. chat name and we can do like a drawing. We can do like a short answer. It's been a while since I did Google forms, but let's see what uh enter your chat name. Okay. Publish. Okay. So, could people go in here? Can you sign up? I just started tapping in yesterday and I got to say, pretty cool what you're doing. Thanks, Drock. Appreciate it. Okay, so you just got GPD5, Aaron. That's cool. Uh ASA said GPD5 is the real leader. I oneshoted a problem my code base that GLM5 was spinning at uh wheels for hours. Okay, that's some optimistic things to hear. Uh looking forward to trying it out. So the form works. Let me see. Okay, so I already have 20 responses. Perfect. So, what we're going to do, we're going to do we're going to have GBD5 create um a Python code to pick a random winner. So, let me share it a few more times. So, let me just uh Oops. So just enter there. You can be like a claw claw claude code uh pro subscription or like a chat GPT pro. It's kind of up to you. Hello Espen to Duck. Scotty Graham. Yeah, we're going to we're going to list it out so everyone can check if they see their names. No, I don't think capital matters. I'm going to I'm going to grab it on uh on uh Google forms. I'm going to grab it here anyway. Too much. So, we have 56 responses. So yeah, if you want to join, just join. We're going to grab um grab everything and we're going to make [ __ ] uh pick a winner and I'm going to make sure it's going to do no duplicates. So no point putting your name twice. We are still working on our Minecraft game here. What about this? We are waiting for our garden. Still crashing. Is it now? How many of you have submitted more than once? Make a terminal calendar where users can uh add view events. Save it to local five. Uh okay, we can do that. Uh Hillary Okay, 77 responses. I can link the form again. Uh so if you want to join, join the giveaway, just sign up here. Yeah, it's not looking too good. So, L, uh, if you're still here, uh, I'm going to give you the link to the answer for the evolution. Okay. So, uh what did you ask about? It was um what was it? a terminal calendar. See if I can see it. We got to look back at our uh Okay, so we're still working on this. It's very slow though. Uh, okay. So, let's try to make this simple calendar. GPT. We can do it here. Open code. Okay. Make a terminal calendar where users can add view events. Save it to a local file and show the current week's agenda. Let's see what happens. Okay, so we are 88 people joining the draw. Let me know if you don't have the link. Yeah, we need to play around more with it. I think it could be interesting. So, GPT5 is now in both Windsurf and VS Code. Max status bar where you can talk and add events to Google calendar. GPD is good so far. Fantastic clone. I think what I think this is maybe a hot take but I think yeah this is not working. Let's say let's say you are a a novice user, right? Yeah. Bar, give me a second. So, you can join here if you want to be a part of the draw. So, let's say you've been using GPT, SH GPT. So, my like my sister, she uses SHA GPT, but she doesn't pick models. She just creates a new chat and she is always like on GP 4.0, but now she's going to be on GP5, right? So, I think she's going to have a better experience now. Hello, Manamorphical Robert R. Nice to see you. Yeah, giveaway link is here if you want to join in. So, I think this is a good for most users. Maybe not. We people in the chat here are power users of AI, right? Could be good for us too. But like most people like my my mom, my sisters, my friends, they kind of use it like casually. I think it's going to be a big upgrade for them. Personally, we'll see. Yeah. Okay. We have passed 100 responses on the giveaway. That's good. Many people maybe have done it two times. Um, so I think we're going to end the stream in 15 minutes. Uh, and let's do the stream in like let's do the drawing in like 10 minutes. Okay, Edgar, if I have time, I might drop by afterwards. Uh, what was the questions? Create an iOS 286 liquid glass UI. I can try that in chat. Yeah, we'll see, Edgar, if I have time. Uh, there are no rules, Bogan. Just write in your name and um just write in your name and you can um win a $20 subscription. Still working on the app. Uh still working on the programming language. I don't think it's going to finish this. Maybe I'll check it tomorrow. I'm just going to let it stand here. So, let's try the liquid glass prompt. I don't have the chat up on my Mac now, so I just I'm just gonna Do it here. Some a few errors, but not too bad. I guess they're rolling it slow and kind of monitoring the GPU usage, I think. Hello, Tim. Uh, said ID. I can't say I haven't. Jacko, nice to see you. Uh, I don't know if you can see previous chats, but uh, we have a giveaway so people can join. So, let's do if you want to be, let's try it out. Did it create an image now? It's pointless. Yeah, the giveaways were usually uh the giveaways were usually for uh some people in the stream. uh they don't really have the subscription money, but uh so I just wanted to do some giveaways, but everyone can join in. Doesn't matter. We got to figure out how to screen chat users. Uh yeah, GPT5 is going to fix the It's going to fix the duplicate. Don't worry. Hello, Michael. Yeah, you can like the stream if you want to. Kind of helps me out. Michael, if you want to join the giveaway, just click on that link. Everyone can join. No problem, Tony. It's going to do an SVG here. Or is it? I don't know. Uh, let me see. It's doing HTML. Is it doing CSS? You tell me. I don't think is this going to be JavaScript. Let's see. So, we did try to create this Minecraft, but it's just running into issues. Just got GPD5. That's cool. GPD5 is out and Horizon beta is still up. Uh doubt that's doubt what she said. I am running uh on open router on open code. I am running GP5. Okay. So let's check our terminal app how to run it. So is this an open R? Okay. So it's just Python weak cal.py. Okay. So, I can add an event. What's this? Okay, so let's see now we are 120 people. It's pretty good. If you got here late, you can join here. uh in the app. Uh is it still old version web page? Uh I got upgraded. Okay, so here's the liquid glass app. Uh cam. What do you think? Pretty interesting. It's not the highest resolution. Just need some tweaks. I don't know what this is for CSS. It's leaked. We can add some add more relection and backlight. Yeah, maybe icons could be white. Yeah, they are barely visible. Yeah, it looks pretty cool. Hello Yanick. Yeah, I don't think we're going to fix our Minecraft game. What about our garden? I think we're going to give up on this. Hello Yan. Uh what is the time here? It's uh yeah 9:51 p.m. Vanishes, what is your thoughts on five so far? Uh we have done some fun things but uh I haven't really done like a deep dive but the pricing uh on the API looks really good. That is pretty exciting. If you kind of compare it, Clo Opus 15 in, 75 out, GPT 5, 125 in, 10 out. Seems like a good deal right? So, I think a lot of people is going to go back to cursor, maybe try to put their API key in, or maybe people are going to go to wind surf. We'll see. Maybe free up some rate limits on cloud code. I'm running it on open router now, but I also have cursor. So, let's do Okay. Yeah, I check this out later. I'm room start Tim. Okay, so this is going to be the last uh complete. It's not sponsored. It's just me. So, this is going to be the last link for the giveaway. We are doing a claude code chhat GPD pro subscription giveaway. Uh yeah, that's about it. So I'm going to prepare GPT5. Let's remove this. Uh the Minecraft game doesn't look too good. Poly market. Okay, Ken, how are you doing? Yeah, let let me know it maybe tomorrow. Okay, so let's do a few more minutes and people can join in if you want to be a part of the draw. Uh then we're going to ask Chat DP team. I uploaded uh CSV file with chat names. create a Python code to pick a random chat name as a winner. Uh remove all dupes from the list and create some antipation before we see who won. Spinning wheel. Okay. Should we try that? That could be pretty cool. create a spinning with that many names with all the usernames. First remove all duplicates. Create a button to spin the wheel. Random winner. Yeah, I don't really care about the email because uh I'm going to give the winner the email afterwards. You can like the stream if you want. Yeah, if I find any duplicates, we're just going to do it again. So, let's try this. Okay. I'm going to give you one more second last time to join. I'm just going to close down some open here. I'm going to close down uh yeah, this is I'm going to close down this. Okay, so the programming language is finished, but uh I don't think we have the time to try it out yet. I'm going to check it out afterwards. Yeah, my quot here is done. Okay, so I'm going to download the CSV file now. Thank you everyone for entering. Yeah, it doesn't matter, Melvin. It's up to you. I'm going to contact the winner on email and then we're going to figure it out. I uploaded a CSV file with chat names. Create a spinning wheel with all the usernames. First, remove all duplicates. Create a button to spin the wheel. A random pick a random a random winner. Let's see first. Yeah l we going to check for duplicates. Hello David Merlin 8 D you were a little bit too late for to join the giveaway but let's see now hope it works now so it's going to be interesting to see what people think of Gypty 5 kind of the sentiment we can go back while we wait for this let's check poly market how it's doing here on um on the prediction markets. So, looks like the market isn't too impressed yet. Uh before the stream, 75% thought OpenAI was going to have the best model by end of August. During the stream, it flipped to Google. So, people are expecting Gemini 3.5 to be better. I don't know. Maybe it can't. Uh Oh, it's a zip file. Oh, I didn't see that. Okay, let's fix that. Yeah, I didn't think of that. Why is there a zip file? Okay, here we go. Okay, should be good now. Okay, so we have 137 Should have added some sound, too. Is he going to do it in Python? That's boring. I wanted like a JavaScript or React or something. I don't want Python. Let's do uh uh react in canvas. Kind of just wanted to do it in the border here. The tips I'm getting on Gifty though is the API is really slow. Uh Mirin, nice stream. I'll be adding G5 to my list of models. Uh okay. Yeah, I can see some people don't need it. But uh have they given any prediction when they will launch Gemini 3 tree 3? I don't think so. With the sound. Okay, let's do with sound then with sound if it's possible. How is the code quality? One punch weekend. Haven't really had time to check it out, but we had someone in stream. They had kind of a prompt they have done before. Why is it doing Python? Or it's just using Python to sort the list, right? I didn't see that. Yeah, it's just going to clean up the list with open. Uh, okay. Is it going to use Python to check for duplicates? Okay, so it it removed seven duplicates. I think if this doesn't work, I'm just going to do it in cloud code. No, that is not right. 30 people looks bad. What? The list had 137. Looks like I accidentally only copied the first 30. Yeah. This Do we trust this? I don't know. I don't know if I trust this accidentally. Yeah. As long as your name is on the list. It's just skipping it, isn't it? Check for dupes. Can I just copy the list here? Use the full list. Never trust confirm twice. Yeah, I don't know why it has to use this. Why does it say include all the remaining names from the dd the dduplicate list? Okay, so let's do the code first. I'm going to copy it and I'm going to add the list. Oh my god. How annoying is this? Fill in all names. Check. I'm going to do cold code. I thought it was going to breathe. This is a new model. It shouldn't take long, right? It has Why does it have to use Python to check for this? I don't get it. It's just just open up the names. Okay. So, I'm going to do it like this. So extract all names separate by uh okay my god. I don't know what kind of remove dupes just go to clo. I agree. I don't want to deal with this. At least we got the list now. Here is the list. Disappointing. Python is the mode you Yeah. But I don't know why does it have to use Python to just remove some names. Everything is in context here. Remove loopy. Okay. This is the last attempt. If not, I'm just going to use code. I'm just going to start up close code here just in case. Okay. So, it looks good now. Yeah, I think we have everyone. Where the hype? Okay, so I think we're good now. It looks like we have all people. Okay, good. It better create a nice uh spinning wheel now. You need to be more specific with your prompt. Yeah, I guess. Didn't I say react? It's just a very simple thing. I don't know what this creating here. Okay, so let's try to run it. Okay, look at all the names. So, dd duplicated. We ended up with 129. So, blame [ __ ] if uh your name is not on here. So, are we ready? It's uh poorly executed. I agree on. Okay. So, good luck everyone. No colors. Okay. So, let's see if it works. If it works, good luck. The winner was Melvin AI. Okay, Melvin, you are the winner. So, send me a email at Yeah, that was pretty good. We got there in the end. So, Melvin, can you confirm that you see the mail, that you have the mail? Okay. Yeah, that was pretty cool. Thanks for showing us how crappy five is so far, Melin. Let me see. Uh, is the spinning wheel just fake? It landed on Stephen. Where is Melvin? Melvin is over here. I'm not going to run it again, though. Here is Melvin. But uh here's the winner. Uh Melvin, can you confirm you got the mail? Okay, great. No, we're not going to do it again. Is it to the right? Okay. Yeah, it's done now. I'm not going to do it twice. Use claw, please. No, I'm pretty happy. I'm pretty happy. Okay, we do one last spin just for fun. If it wasn't for Melvin, it would have been Let me turn on the sound of it. Runner up, Venicious Ato 1047. Yeah. So, too bad, but uh I'm happy. Okay, guys. Thank you everyone for tuning in. So, definitely go check out Chip 5. Uh I'm going to explore it more this weekend. I might even do like a stream tomorrow. So, yeah, I think so far pretty good for UI stuff. Need to test it out more. uh pretty good for everyone else that is not early adapters like my sister, my mom, they're going to really enjoy this, I think. Uh yeah, they're going to try it out. So, thank you everyone for hanging out. Uh it's really fun not to just watch it alone. Uh thanks Unreal, Jackass, Matias, AI, Code L Michael Drock Kevin Jesse Jup, thanks for the likes and Ms. Can you share the code for the wheel? Uh, yeah, I might put it up. You should be able to recreate this. Merlin, Jake, Tim, Tech Friend, Tech Friend, I did send you an email. Let me know if you got it. Just uh let me know. Uh Ken, thank you. One Punch Van. And yeah, have a nice day. Play around with GT5 and I'll hopefully see you again uh next time. Bye. Bye.