# NEW Gemini 3.1 Flash-Lite Update + ChatGPT 5.3 Instant

## Метаданные

- **Канал:** Julian Goldie SEO
- **YouTube:** https://www.youtube.com/watch?v=nxH3rd9LqXQ
- **Дата:** 04.03.2026
- **Длительность:** 32:40
- **Просмотры:** 1,308
- **Источник:** https://ekstraktznaniy.ru/video/10764

## Описание

Want to make money and save time with AI? Join here: https://www.skool.com/ai-profit-lab-7462/about

Get a FREE AI Course + Community + 1,000 AI Agents + video notes + links to the tools 👉 https://www.skool.com/ai-seo-with-julian-goldie-1553/about

In this livestream, I cover the brand new Gemini 3.1 Flash-Lite update and break down ChatGPT 5.3 Instant. If you want to stay up to date on the latest AI model releases, this is the stream for you.

## Транскрипт

### Segment 1 (00:00 - 05:00) []

Chat GPT just dropped 5. 3 instant. So, Open Eye just quietly rolled out GPT 5. 3 instant to every single Chat GPT user on the planet. It's free, no upgrade needed. And this isn't just a Vibe Code update. According to the stats, hallucinations are down almost 27%. Um, it also feels a lot less cringe. We're going to be testing out today. The web search is improved, too. and it's also giving better answers. So, this is the chat boutique people wanted from the beginning. Should be a lot faster when it replies as well. And today, we're testing it live together to see if it actually delivers because everyone's hyping this up right now, but does it actually feel different? Is it actually smarter in reality or is this just opening doing some sort of damage control after the last updates? Let's check it out. Now, stick with me because by the end of this video, you're going to know exactly what changed, whether it's actually worth using, and one thing most people are completely sleeping on when it comes to this update. So, we're going to check it out now. You see that it's inside chat GPT like so, and we can start testing this out. So, if we go inside here, um, it should be a lot faster. We can probably compare it side by side in a second as well. So, I'm just going to say, uh, something quick like this just to see how it performs. It's really fast to reply to be fair. you was ste still using those m dashes which to this day I don't understand why we're doing that. Um but it is super fast to reply on as well. And according to the tweet announcement this is by Nick Turley who's VIP and head of the chat GPT app. He said that we've heard the feedback on 5. 2 sometimes too cautious too many caveats and conversations that didn't flow as naturally as they should. 5. 3 instant tackles that with fewer unnecessary refusals, fewer defensive disclaimers and answers that get to the point more directly. So, if we actually have a look at the benchmarks here and we site itself. Let's check this out. This is literally it's just been announced and they said GPT 5. 3 delivers more accurate answers, richer and better contextualized results when searching the web and reduces unnecessary uh dead ends caveats and hopefully declarative phrasing that can interrupt the flow of conversation [snorts] and then it focuses on the like the tone relevance and conversation flow. Let's have a look at how it's being used. So you can see like sideby-side comparisons right here. So tends to split up the text creates like shorter responses and everything else. We actually have a look at that prompt here. It says can you help me with the trajectory of calculations for really long distance archery scenario. So let's try that out inside chat GPT here. What we can actually do is just split screen this. So if we go to another tab of chat GPT I'm going to select 5. 2 2 and we can just compare how they perform side by side. So we use GT2 instant versus GT3. There we go. So we can compare them side by side. And then we'll put the same prompt into both of these. Also, if you look at this 5. 2 instant doesn't have a thinking option, whereas it looks like you get a thinking option with uh let's have a look here. Instant. Oh, no. They're both non-thinking models. It just selected 5. 2 thinking for some reason. team says, "Hi, are you real or Avatar? What do you think, mate? " Um, so if we have a look here side by side, this is like a more straight to the point response, whereas this is like a longer list. Which one feels more natural? This one, I would say this one just gets a lot more to the point, right? it flows better. And this is GPT 5. 3 instance. So it does seem like it's giving better answers in the conversational style. Um, also if you look at the way that it's formatted, this answer here looks way nicer. So look at the way that it's added the formulas and everything else.

### Segment 2 (05:00 - 10:00) [5:00]

Whereas with exactly the same prompt on chat GPT 5. 2 instant, the old one, it just looks less interesting, a bit more boring, a bit more sort of too logical, and it doesn't feel as sentient as the left hand side. So you definitely get a better answer on the left hand side and on the right hand side it doesn't look so nice. So it's definitely an improvement in the right direction. So let's come on to more examples and what it talks about. So more useful well synthesized answers when using the web. So it said GP5. 3 instant also improves the quality of answers when information comes from the web. it more effectively balances out what it finds online blah blah and it's less likely to overindex on web results which could lead to long list of links or loosely connected information. So here's another example. This is 5. 2 to the old version versus the new and they said what was the biggest signing of the last B baseball off season and the old one it kind of writes like a long blog post. The right hand side looks a lot more sort of um nicer formatted and that. So let's have a look and test this out ourselves. We'll [snorts] put this inside the same chat. This was way faster. By the way, GT5. 3 is a lot faster on the responses. But I say like the I mean the results look kind of better on 5. 2. 2 instant. You see how you get like the news links here, which I think looks nicer, nicer emojis. You see how it's got these emojis on the answers? But this one is more direct and straight to the point. It's more punchy. It says smoother, more to the point conversational style. I've definitely noticed that. So, a lot of people said that GT5. 2 instance tone could sometimes feel cringe, whereas, for example, with this update, it's just a lot more sort of direct and punchy and a bit smoother when it comes to speaking with it directly. Honestly, I think I did cancel chat GBT after the last couple of updates because I felt like it's just nowhere near as good as something like Claude. Uh so type this in. And you can see here that this gives a more direct response than this one. That is a brutal question. Let's see what else we got here. More reliably accurate responses. So I think I heard Yeah. So on the hero rates stakes evaluations GT5. 3 instant reduces hallucination rates by 26. 8% and when using the web and 19. 7% when only relying on its internal knowledge right so it's a big improvement in terms of reducing hallucinations which is great I have noticed that with claude as well. Claude hallucinates way less uh than other stuff as well. and then you can see a big difference here. So someone said write a short poem and then 5. 2 two sort of says here's a poem blah blah whereas GPT 5. 3 instant doesn't even explain it just gives the poem directly inside the chat which I think people would want right it saves time let's talk about limitations so 5. 3 instant makes more meaningful progress on everyday usability um but it's not as good on non-English languages so Japanese and Korean not so good and then tone as well so it's smoother but I think they're going to improve it the other thing that chat recently announced let me find So, they've actually said the 5. 4 is coming sooner than you think, which is interesting. So, I think they're going to drop 5. 4 very soon. Maybe this week or next week. It's got to be coming sooner. Um, they've definitely hinted at it on their X account. So, it's just a matter of time on that as well, which will be interesting to test out. So overall big improvement I would say with 5. 3. It's faster, smoother, more direct, more punchy. You're getting less

### Segment 3 (10:00 - 15:00) [10:00]

cringe answers and um I wouldn't say this is particularly useful for like coding and stuff like that, but it's more for like quick answers, web searches. That's what it's designed for, and it's better to give you a better tone. So overall, it does outperform 5. 2. Am I saying it's better than Claude or anywhere near the same range as Claude? Absolutely not. I would say Claude is um you know Claude is just miles ahead of anyone at this point. So that's basically it. Now if you want to get all the video notes from today along with a 30-day plan for mastering this plus 100 prompts that you can actually test as you can see right here for SEO sales business operations etc with GPT 5. 3. You can find that in the AI profit boardroom along with a 6-hour course on openclaw and daily updates on what's working and what's not. We also inside the community do weekly updates. So I condense my 60 hours of research per week with AI and I condense it into a quick update that's like here's the good stuff and here's what you can ignore based on what's working. So you can see here basically I do 60 hours of research. You can skip straight to what matters and I break down exactly what's good and what's not inside the AR profit board and community as well. We have 2,500 uh sorry 2,600 members now. And um basically what this is about is just helping you save time, scale, and grow with AI automation, right? So if you want to learn how to implement this stuff in your business, we have a supportive, helpful community that's focused on helping you grow with AI automation. You can ask questions, share your wins, get help and support, see what other people are building as well, which is pretty cool. We have a daily accountability group as well, which means that you can stay focused on your goals and avoid information overload. The same with the weekly AI updates, right? is like I cut through the noise and I give you the most important updates so that you don't feel overwhelmed. You can be like right here's what's good and here was what you can ignore, right? And that saves a lot of time. Also, inside the calendar, we have four weekly coaching calls so you can stay um so you can jump on live coaching calls, get help, get support, meet the community, etc. And we have over 133 pages of testimonials from uh people inside our school groups, right? So you can see all the amazing testimonials that we've got. Uh people winning. This is 133 pages of this stuff, right? So people just getting awesome stuff, which is great, you know, people getting results and that's what we want from this community. Also inside the air profit boardroom inside the classroom here you can see the for example we show you how to go from beginner to expert with AI automation and you'll learn how to build your first AI agent in under 5 minutes. Additionally you can get my best playbook. So these are the business processes that I use and the automations I personally use for my business. Right? So for example how I automate ads newsletters shorts Instagram AI avatar videos all those systems are inside this section. Additionally, if you want to get daily updates on the most important stuff, you can do that inside this section. So, we've got a full six-hour course on OpenClaw and another three-hour course right here with step-by-step guides. But you can also see that every single day we add new guides based on the most important updates I've tested. Right? So, you get video updates. I test all this stuff. I show you how it works. We give you the full notes on exactly how to implement it. And the cool thing about this is that we also give you step-by-step guides and video tutorials so you can implement this, right? and I just look at all the new updates that came out today and then I focus on the most important ones that would help you the most. Inside the classroom too, you can learn how to get more agency clients with the agency course. You can watch back the coaching calls as well. You can learn how to rank number one with AI SEO inside Google and other AI search engines. And also you can learn how to grow a YouTube channel with AI um using the YouTube 6E road map based on what's working for me. So that's all inside the AI profit board. Feel free to get it link in the comments description or just go to the aiprofitbot. com. Javam says, "Did GPT 5. 3 instant roll out to you? It didn't roll out to me yet. " Yes, as you've seen today, um I've already got it. I think if you don't have access, you could probably go to the playground on chat GPT and then you could start using it there. Anyway, uh also it's available to test on LM arena. So if you don't have access to it directly inside chat, you can easily find it in other places. If we actually go to for example uh open router, you can get access to it there. Right? So if you don't have access to it directly inside the chat, just go to open router and you can try it out inside the chat if you really want to try it. Right? Um one thing to note here is it's got a pretty low context window. So this is definitely not designed for like super complex tasks. You can see here 128,000 context window, right? Which is super small. That's even smaller than um Alibaba's local free models that are running locally on your laptop with a quen 3. 5 that came out yesterday. Right. So the main point here is like the context window is super low. So this is just really designed for like short quick answers and questions.

### Segment 4 (15:00 - 20:00) [15:00]

Google just quietly dropped something massive. So, Gemini 3. 1 flashlight, brand new, just released. Here's what makes this insane. This is a budget AI model, a small, cheap, fast model, but it just outperformed Google's bigger, more expensive models on reasoning tests. It's basically a budget car that just beat a Ferrari. And that's what happened here. And the best part is you can use it for free right now. Let's test it out. Um, and we're going to get straight into this. So, you can see Gemini 3. 1 flash light built for intelligence at scale. It was just announced today, which means that you get basically a much cheaper, faster version of Gemini. And you can actually use it for free as well, but if you're coding with the API, then it'll be a lot cheaper, but it's faster than ever as well. This is like a smaller model, but it's pretty powerful. And I've tested, for example, Gemini Flash 3. 0 in previous tests. Now, this is 3. 1 Flash Light. As you can see right here, it outperforms 2. 5 flash with a two and a half faster time to first answer token and a 45% increase in output speed according to the artificial analysis benchmark. Right? So this is very low latency and if you need high frequency workflows, aka you need to move fast and build fast, then this is great. The other thing that this is noted for is responsive realtime experiences, right? So if you look at for example GPT 5. 3 they just released GPT 5. 3 instant. If you look at for example claw code they just released a fast model. Everybody wants faster models when it comes to building with AI. Um it's no different with 3. 1 flash light right. So we can see how it performs on benchmarks over here. So in terms of speed and cost and output speed right you can see that Gemini 3. 1 flash light is absolutely flying ahead of all the other models right. So it scores 363 um versus for example Gemini 2. 5 flash GT5 mini claude haiku gro 4. 1 fast they're not even in the race at this point right as you can see and then in terms of price GP 3. 1 flash light is a lot cheaper than look at the cost of for example claw 4. 5 haiku right pretty expensive $5 versus 1. 5 for Gemini 3. 1 flash uh you can see an example of the answers in real time. So this is 2. 5 flash versus Gemini 3. 1 and this just gives much faster answers. So 2% complete versus 0% complete. And then also we can see the full benchmarks here right. So if we look at all the fast models so GP5 mini haiku g 4. 1 fast etc on pretty much all the benchmarks Gemini 3. 1 flashlight is winning. like GP, GPQA diamond, humanity's last exam, etc. Um, it's only Grock 4. 1 fast that is still beating Gemini 3. 1 Flashlight on humanity's last exam. In terms of output speed, it's much higher than almost everything apart from 2. 5 flashlight. And in terms of the price, it's actually quite a lot higher than 2. 5, but uh cheaper than 2. 5 flash, but more a lot more expensive than flashlight, right? [snorts] And then you can see the examples right here. So, it's also got standard thinking levels as well. And you can use this inside AI studio, too. I think this will be available inside Gemini now as well. We can check today and see how it performs. And then we can run it through a few tests and see how it's doing so far. Shivam says, "How is GPT 5. 3 instant versus 5. 2? " Check out the previous video on that. We've already covered it. Um, but yeah, 5. 3 is a lot better. If you've got questions, by the way, just make sure you leave them inside the chat and I will answer them at the end. So, you can see here this is the announcement and if we want to get access, can we get access inside the chat? Let's have a look here. We'll see what API it's using. I don't know if it's fully rolled out to

### Segment 5 (20:00 - 25:00) [20:00]

Gemini yet. — [clears throat] — So, it's Gemini 3 Flash still inside. Yeah, as you can see here, it's still using Gemini 3 Flash inside Gemini. So, the new update is not being rolled out there yet. Let's have a look inside AI Studio. So, if we go back to the dashboard, we can go to the playground and then we just need to make sure that we select flash. Let's find it. There we go. All right. So, what you want to do if you want to get access to this is you would go to the playground here. Then you're going to click on this model, type in flash, and scroll down to Gemini 3. 1 flash light preview. That's how you can get access to this model. Doesn't seem to be rolled out on my version of Gemini yet. So, from there, what we can do is we can go inside here and just test out in terms of how it performs. So, We'll test out with a very simple prompt for creating space invaders. See how that performs. Wow, that is fast. This it literally just coded out Space Invaders in like one click. I don't know if the output's going to be any good or not, but we can test it out. So, if we click on show preview, boom. Look at that. That's crazy. It's a little bit buggy on the game, but the fact that it coded out so fast is actually wild. Let's test out the same prompt on GPT 5. 3 instant and we'll see if it can do that. Um, but that's very impressive. The speed of that was just outrageous considering that actually works. So, if we go to 5. 3 instant, we'll try and create the same thing. We're going to turn on the canvas. See if it can do it. So, I mean, like 5. 3 instant is super fast, too, but it doesn't seem anywhere near as fast as 5. 3 flash. Uh, sorry, as Gemini 3. 1 flash. Yeah, see, it's taking quite a while. We can come back to that later. Oh, there we go. We got the preview. All right, let's try this out. Uh 5. 3 has done a decent job as well. You know, it's not quite as fast as Gemini 3. 1 flash light, but it's decent. So, let's test out another prompt here. Try and create a Flappy Bird game. standard change to Craig says, "How are most people handling OpenClaw installation? " Yeah. So, the way that I do it is I just install it locally and then it's running from the cloud once I connected to Telegram. Um, if you want to use a VPS, don't use like a you know your common sort of VPS methods for OpenClaw. Really what you want to do is you want to set up if you if you're insistent on using a VPS, right? Use molt worker. Molt worker is a lot more secure um runs in a sandbox. You have to approve anyone's login before they can start using your open claw. And that's just a lot better, right? That is a lot better than using like a VPS where someone can um you know get access to the gateway and then you know they can get access to open claw which would be no bueno. So let's have a look here and see how Flappy Bird performs. So, here's Flappy Bird. There's no obstacles to be fair. I do feel like with flash, Gemini 3. 1 flash light, it'll give you a quick answer, but it doesn't think about the answer much, right? From what I've seen so far. Now, I wonder if you can actually change the thinking level. So, you can change the thinking level inside AI Studio, and that'll probably give you better results, right? So, if we take that, we're going to open up a new chat here. We're going to switch the thinking level from minimal, which is as low as it possibly can be, to something like medium. We'll just see how that performs. Uh, also something to note here is I tried to check inside anti-gravity to see if Gemini 3. 1 flashlight was released yet, and it's not there yet, right? So, still the old version of Gemini 3. 0 flash is inside anti-gravity, but I couldn't see a Gemini 3. 1 flash. Just something to note. Muhammad says, "What's going on? " on.

### Segment 6 (25:00 - 30:00) [25:00]

So, we are testing Gemini 3. 1 Flash Light today. Um, let's show the preview of this. Yeah. So, it's not bad, but it's when it creates stuff, it's a little bit buggy as you can see. Like, it just it doesn't feel that smooth. Um, but it does give you super fast answers. So, it's nowhere near like the same level as something like Gemini uh 3. 0 Pro. We'll let that go. All right. Um, but it does give you extremely fast answers. Even on like medium thinking level, it can do a good job. Just to put that in perspective, if we switch this over to something like for example uh Gemini 3. 0 Pro, so we can use 3. 1 Pro, right? This is like a much better um more powerful API. And we keep the thinking level on high just to show you how better the outputs will be, how much better they'll be with the Pro model. We'll run this example, right? And again, you can get access to these models for free inside Google AI Studio, right? You just use the playground and then you can use all this stuff for free. Now, just to recap whilst we're doing that, whilst waiting for this to code out, just to recap on the notes, you know, what is Gemini 3. 1 Flashlight and why should you care? So, Google released Gemini 3. 1 Flashlight on March the 3rd today, right? Um, this is fresh off the press and it's now in live preview on Google AI Studio, which is free to access. Or if you're on an enterprise team account, then you can use Vertex AI. Right. Here's what makes it different. It is two and a half times faster than Gemini 2. 5 Flash, the model that it replaces. It is 45% faster at generating outputs than its predecessor. It outperforms larger, more expensive Gemini models from the previous generation on key reasoning benchmarks. It scored 86. 9% on GPQA diamond, a test for designed for expert level reasoning whilst costing a fraction of bigger models. I just want to double check on the context window. Let's have a look here. Yeah, so it does have a this is pretty insane. Million context window, right? That is huge. That's as big as it can be currently. Um, so it has a 2 million token context window, meaning it can read and process enormous amounts of text in one go, much bigger than, for example, GBT 5. 3 Instant, which has a 128,000 context window. It supports text, image, video, and audio input. It comes with builtin thinking level, so you can literally dial in how hard it thinks before it answers. So if you're just doing fast tasks, you would use minimal, high for complex reasoning. And that's brand new as well, right? So if you want to change the thinking level, you just do it on the right hand side over there inside AI Studio. Now, if we preview the new version, so this is like the new version of um of Flappy Bird that we built inside AI Studio with Gemini 3. 0 Pro. And you can just see how much better that looks versus the older version um using Gemini 3. 1 Flash Light. Right. So Gemini Pro is always going to out compete the flash models. Um, so if you want the best possible outputs, then of course you would use something like Gemini 3 Pro. But if you want something that's fast, that's real time, that's cheaper, then you would go with Gemini um 3. 1 Flash Light. It's the most cost efficient model in the entire Gemini 3 series and you can access it today at a uh studio aistudio. google. com. So, thanks so much for watching. We've tested out. We've seen how it performs. It is pretty impressive as a model. Super fast. Um, performs very well on the benchmarks and gives good outputs in real life, too. We actually got a 30-day plan if you want to master this inside the AI profit board and video notes from today. And, you know, just in terms of how this works. So, some people want to say, I'm not technical enough to use AI tools. If you can type a sentence, you can use Gemini 3. 1 Flash, right? There's no code, no installation, no developer knowledge needed. You just go to AI Studio and type in a prompt as you've seen today. Other people are going to say, "Well, AI outputs are generic and useless for my business. " Honestly, the more specific the prompt, output, right? So, if you're using something like Gemini 3. 1 Flash, you can easily create like amazing content, amazing tweets, amazing blog posts, etc. As long as you're very specific with the prompt. And um yeah, we've also got 100 prompts you can test out with this. As you can see, split by category. So, for example, lead generation prompts. We've got prompts for, for example, content creation. We have prompts for business strategy, etc. And that's all inside the AI profit

### Segment 7 (30:00 - 32:00) [30:00]

boardroom Gemini 3. 1 flash light notes today. Now, this is my AI automation community with 2,600 amazing members. Um, inside here we have a six-hour course on openclaw plus a 3-hour course. We update this daily with new current guides as you can see. So, yesterday we added uh six guides inside the AR profit boardroom. We add more every single day and we just look at all the updates that just got released like Gemini 3. 1 flash light just got released a few hours ago and we've already tested it and shown you how it works today and inside here the goal here for everyone who joins is like to save time scale and grow their business with AI automation right so inside the community you can ask questions get help get support whenever you want to have weekly updates where we basically break down 60 hours of my research and then show you what are the best things and what things can you ignore. as well to save you time. You also get a daily accountability group which allows you to stay focused and avoid information overload. And then also inside the calendar here, you can jump on weekly video calls where you can get AI automation advice. You can share your screen, ask questions if you ever get stuck, you can get help and support in real time. And we actually have over 133 pages of testimonials from people winning with this stuff. Right? So this is pretty much an infinite scroll of people just getting awesome results with the AR profit boardroom and you can see all the amazing testimonials on screen right now. Right. So if you want to get results to this um you know if you want to join a community that's learning growing supporting raising each other up uh full of positivity then you can check this out. Also inside the map here, we have this map where you can basically zoom in on your local location and DM people and meet people in your local area doing the same things, right? And also inside the classroom here, you get access to all my best training. So if you want to go from like beginner to expert with AI automation and learn how to build your first AI agent in under five minutes, you can do that right here. If you want to get the best AI automations that I personally use for my business, then you can check out these sections. They have video tutorials and full step-by-step guides on how I automate, for example, Twitter and newsletters and emails and shorts and Instagram and AI avatar videos, etc. If you want to learn and get the daily updates on AI and what's important, then you can check out this section. We updated daily. It's probably the most active community out there when it comes to the new updates. And then also, we have a section here for getting more clients with the agency course. You can watch about the coaching course here. And you can also, for example, learn how to rank number one with AI SEO and grow a YouTube channel based on what's working for me. So that's all inside the AR profit boardroom link in the comments description or go to the arprofitboard. com.