OpenAI“最强模型”5

当地时间周四（8月7日），OpenAI官方正式推出了备受外界期待的、性能更强的新一代人工智能模型GPT-5。根据OpenAI介绍，作为OpenAI的“最强大模型”，GPT-5在三个关键领域实现了显著提升，跟着发布会来了解一下。

Good morning.

早上好。

32 months ago, we launched ChatGPT.

32个月前，我们发布了ChatGPT。

And since then, it has become the default way that people use AI.

从那时起，它就成了大家使用人工智能的首选方式。

In that first week, a million people tried it out.

首周就有一百万用户体验了它。

And we thought that was pretty incredible.

我们觉得这非常令人震撼。

But now, about 700 million people use ChatGPT every week and increasingly rely on it to work, to learn, for advice, to create, and much more.

现在，每周约有7亿人使用ChatGPT，越来越依赖它来工作、学习、咨询建议、创作，及更多用途。

Today, finally, we're launching GPT-5.

今天，我们终于发布了GPT-5。

GPT-5 is a major upgrade over GPT-4 and a significant step along our path to AGI.

GPT-5是对GPT-4的重大升级，是我们迈向通用人工智能的重要一步。

Now, today we're going to show you some incredible demos, we'll talk about some performance metrics, but the important point is this.

今天，我们将展示一些精彩的演示，分享一些性能指标，但最重要的是：

We think you will love using GPT-5 much more than any previous AI.

我们相信，你会比以往任何AI都更喜欢使用GPT-5。

It is useful, it is smart, it is fast, and it's intuitive.

它实用、智能、快速且操作直观。

But it's not only asking now.

它现在不仅能回答问题。

GPT-5 can also do stuff for you.

GPT-5还能帮你完成各种任务。

It can write an entire computer program from scratch to help you with whatever you'd like.

它能从零开始写完整程序，帮助你实现任何需求。

And we think this idea of software on demand is going to be one of the defining characteristics of the GPT-5 era.

我们认为“按需生成软件”将成为GPT-5时代的核心特征之一。

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.

到目前为止，直至目前，我们的用户只能在标准GPT的快速响应与推理模型的缓慢而深思熟虑的响应之间做出选择。

But GPT-5, it eliminates this choice.

但GPT-5消除了这种取舍。

It aims to think just the perfect amount to give you the perfect answer.

它会恰到好处地思考，给出最完美的答案。

Today, we're going to show a series of demos in coding, in writing, in learning, and in health.

今天，我们将展示编程、写作、学习和健康领域的一系列演示。

But GPT-5 isn't limited to these domains.

但GPT-5的应用远不止这些领域。

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.

它在需要深度推理或专业知识的领域非常有用，比如数学、物理，甚至法律。

And the exciting thing is we're excited to make this available to everyone, even to our free tier.

令人兴奋的是，我们将向所有用户开放，包括免费用户。

The best part is that we're bringing this frontier intelligence to all users.

最棒的是，我们把最前沿的智能带给了所有人。

GPT-5 is rolling out today for free plus pro and team users, and next week we'll roll it out to enterprise and edu.

GPT-5今日向免费用户、Plus用户、专业订阅用户和团队用户推出，下周将推广到企业和教育用户。

For the first time, our most advanced model will be available to the free tier.

这是我们首次将最先进模型开放给免费用户。

Free users will start with GPT-5, and when they hit their limit, they'll transition to GPT-5 Mini, a smaller but still highly capable model.

免费用户将先使用GPT-5，达到使用上限后，会自动切换到GPT-5 Mini，体积更小但依然非常强大。

It actually outperforms O3 on many dimensions.

它在多个方面超越了O3模型。

Plus users will still have significantly higher usage than free users, and our pro subscribers will get unlimited GPT-5,

Plus用户拥有远高于免费用户的使用额度，专业订阅用户可无限制使用GPT-5，

along with GPT-5 Pro, extended thinking for even more detailed and reliable responses when you just need that extra depth.

获得GPT-5 Pro版本，在需要更深层次回答时，提供更详尽且可靠的扩展思考。

Suppose your kid is in middle school physics and they want to learn about Bernoulli effect.

假如你的孩子正在学初中物理，想了解伯努利效应。

They need your help with their homework and you might be like,

他们做作业遇到难题，你可能会说：

Wait, I might need some help with that too.

“等等，我自己也得搞懂。”

So you could ask GPT-5, give me a quick refresher on the Bernoulli effect and why airplanes are the shape they are.

这时你可以问GPT-5：“帮我快速复习一下伯努利效应，还有飞机为何是那种形状。”

Since this is a pretty straightforward prompt, GPT-5 actually doesn't need extra time to think about it, and it answers right away.

由于问题很直接，GPT-5无需额外思考时间，可以立刻给出答案。

But it still gives me a high-quality answer, and it explains the concept clearly.

但回答依旧高质量，讲解清晰明了。

So here it says, Bernoulli effect means faster-moving fluid has lower pressure, and slow-moving fluid has higher pressure.

它说，伯努利效应指流速快的流体压力低，流速慢的流体压力高。

So to make this even more helpful, I'm going to ask GPT-5 to create a moving demo to illustrate this.

为了更直观，我让GPT-5制作了一个动态演示来说明这个原理。

So I could ask, explain this in detail and create a moving SVG in the canvas tool to show me.

我让它详细解释，并用可画工具创建动态SVG给我看。

This is a pretty complex task because now GPT-5 actually needs to build the visual.

这是个比较复杂的任务，因为GPT-5需要自己生成视觉内容。

Therefore, GPT-5 takes a moment to think through the answer so you can come back with something more comprehensive and accurate.

因此，GPT-5会花点时间深思，给你更全面准确的结果。

So you see that GPT-5 has already written more than 200 lines of code already.

你会看到GPT-5已经写了200多行代码。

And while the model is thinking, you can also tap here to expand the chain of thought to actually see what's going on under the hood.

模型思考时，你还可以点击这里展开思路链，查看它的内部逻辑。

For example, the GPT-5 was thinking about, oh, the user wants a moving SVG visualization in Canvas.

比如GPT-5在想：“用户需要可画里动态SVG可视化。”

I actually need to create HTML code to do that.

我得写HTML代码实现这个功能。

Okay, so it's already done.

好了，代码写完了。

So it looks like TryGPT just finished like 300 or near 400 lines of code in two minutes.

看起来TryGPT在两分钟内写了将近300到400行代码。

So let's see if the code can actually run.

接下来测试代码能否运行。

Okay.

好的。

Nice.

很棒。

Yeah.

是的。

So with just a simple prompt, GPT-5 created this interactive and engaging demo that I can actually play with.

仅用一个简单提示，GPT-5就生成了这个我可以互动体验的演示。

So I can actually change the air speed here to see how the lift and the pressure change accordingly.

我能调节风速，观察升力和压力的变化。

I can also tweak the angle of attack. to see if my plane will actually fly or crash.

还能调整攻角，看看飞机是飞起来还是坠毁。

I hope not.

希望别坠毁。

Yeah.

是的。

I will actually try to build something that I would find useful, which is building a web app for my partner to learn how to speak French so that she can better communicate with my family.

我还想试着做点对我有用的东西，帮伴侣做一个法语学习网页应用，方便她更好地和我的家人交流。

So, here I have a prompt.

这是我的提示词。

I will execute it.

我马上执行。

It asks exactly what I just said.

内容和我说的一模一样。

Please build a web app for my partner to learn French.

“请帮我做一个法语学习网页应用。”

One thing to note is that GPT-5, just like many of our other models,

需要注意的是，GPT-5和我们其他模型一样，

has a lot of diversity in its answers.

回答非常多样化。

So what I like doing, especially when you do this type of vibe coding, is to take this message and ask it multiple times to GPT-5, and then you can decide which one you prefer.

所以我喜欢用同一个提示，多次让GPT-5生成答案，然后挑出我最满意的版本。

So I'm gonna open a few tabs. I'm just gonna paste. There.

我会开几个标签页，粘贴提示词，好了。

Great.

很好。

So while it's working on it, let's read through exactly the prompt I wrote. Create a beautiful and highly interactive web app for my partner, an English speaker, to learn French.

等待时，先仔细看看我的提示词：“帮我做一个漂亮且高度互动的法语学习网页应用，针对我的英语母语伴侣，让她学法语。”

And then I gave a little bit more details. Track her daily progress. Use a highly engaging theme. Oh, it's already working.

我还补充了一些细节，每天学习进度跟踪，用高互动性主题，哦，已经开始运行了。

I'm going to... put it on the side for now, use a highly engaging theme, include a variety of activities like flashcards and quizzes that she can interact with.

我先放一边……用一个很吸引人的主题，包含多种互动活动，如闪卡和测验。

And then to make it even more fun for her, I actually asked GPT-5 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 replace the snake with a mouse. and the apples with cheese.

为了增加趣味，我让GPT-5嵌入了一个教育小游戏，基于经典贪吃蛇游戏，但加入了法式元素，把蛇换成老鼠，把苹果换成奶酪。

So I can simply press run code.

这样我只要按“运行代码”。

So I'll do that and cross my fingers.

按下去，祈祷一切顺利。

Whoa.

哇。

Oh, nice.

太棒了！

Voila.

瞧！

So we have a nice website.

我们有了一个漂亮的网站。

Name is Midnight in Paris.

名字叫“午夜巴黎”，

Oh, I love that.

我喜欢这个名字。

So romantic.

真浪漫。

We also see a few tabs, flashcards, quiz, and mouse and cheese, exactly like I asked for.

还有几个标签页：闪卡、测验，还有老鼠和奶酪小游戏，完全符合我的要求。

I will play that.

我要试试这个游戏。

So this says Le Chat, which is the cat, sorry.

这儿显示“Le Chat”，就是“猫”的意思，抱歉刚才没注意。

Le Chat.

“Le Chat”。

Well, that's pretty good pronunciation.

发音还挺标准的。

What does that mean?

这是什么意思？

The cat.

就是“猫”。

So I can review and check if GPT-5 is correct.

我可以复查，确认GPT-5给的是正确答案。

It is.

确实没错。

So if I press next, oh, and I don't know if you saw, I think it actually updated the progress bar, which is exactly what I had asked for.

如果我按下一步，不知道你们有没有注意到，进度条确实更新了，正是我想要的功能。

Let's check the quiz.

我们来测试一下测验功能。

Here is the word non, which is no.

这里有个单词“non”，意思是“不”。

So if I press on it, bien joué, which means congrats, and it updated the progress bar again.

点开后显示“bien joué”（干得好），进度条又更新了。

And let's check the mouse and cheese tab.

接着看看“老鼠和奶酪”标签页。

Okay, that seems like a mouse.

这确实是一只老鼠，

Here's the cheese.

还有奶酪。

I'm going to try to play it.

我准备试玩一下。

I can't promise I'm going to be good at it.

不保证我玩得好。

Okay, seems to be working.

看起来运行正常。

Indeed, just when I eat the cheese, it gives me a new French word.

果然，吃到奶酪时会出现一个新的法语单词。

It's actually super complicated and I already lost.

这游戏挺复杂的，我已经输了。

So we've been steadily improving voice over the past year to make it more useful for everyone.

过去一年，我们不断改进语音功能，让它更实用。

First, it sounds incredibly natural, just like you're talking to a real person.

第一，它听起来非常自然，就像和真人对话一样。

Second, we've added video so that it sees what you see while chatting with you.

第二，我们新增了视频功能，可以看到你所看到的内容，同时和你聊天。

Third, it also translates between languages consistently and smoothly across tones.

第三，跨语言翻译时能保持语调流畅自然。

But today, we're doing something very special.

不过今天，我们要带来特别的新功能。

where we're bringing our best voice experience to everyone.

我们将最棒的语音体验开放给所有用户。

Free users can now chat for hours, while paid subscribers can have nearly unlimited access.

免费用户可以畅聊数小时，付费用户几乎无限制使用。

And voice is also available in custom GPT.

自定义GPT同样支持语音功能。

Plus, subscribers now can custom tailor the voice experience exactly to their need.

Plus订阅用户还能根据需求完全定制语音体验。

It will follow your instruction closely.

语音会严格按照你的指令执行。

Let's try something fun.

来试点有趣的。

Excited.

好期待。

So I can ask the voice model to give me a comprehensive answer,

我可以让语音模型给我详细的回答，

a concise one, or even just a single word.

也可以简洁回答，甚至只用一个词。

Hey, Chad, could you only answer to me in one word, please, from now on?

“Chad，从现在起请只用单词回答我，好吗？”

Absolutely.

“没问题。”

Okay.

好的。

Describe the plot of the book Pride and Prejudice.

“简述《傲慢与偏见》的故事。”

relationships.

“爱情。”

Okay, that's true.

嗯，确实如此。

Yeah.

对。

Give me a piece of wisdom.

“给我一句人生智慧。”

Patience.

“耐心。”

So that's the word of wisdom from our voice model.

这就是语音模型给出的智慧之词。

And I think our model is trying to say, thank you for your patience waiting for GPT-5.

我觉得它的意思是：感谢你耐心等待GPT-5的到来。

So I have a younger cousin, and I want to make a game for her.

我有个小表妹，想给她做个游戏。

So I want to make a 3D game that incorporates a castle.

我想做一个包含城堡的3D游戏。

So you can see my prompt.

这是我的提示词，你看见了吗？

I'll just kick this off.

我马上开始执行。

Oh, sorry.

抱歉。

It's always the non-AR part.

总是非AR的地方出问题。

Yeah, exactly.

没错。

Yeah, okay, so you can see my prompt, create a beautiful castle.

我的提示词是：创建一个漂亮的城堡。

I've included some details like we want people patrolling the walls, some movement, horses.

我还加了细节，比如城墙上有巡逻的人，有动作，还有马匹。

And I want a mini game where I can pop balloons by clicking on them.

还要一个点气球就能戳破的小游戏。

And this should make a sound effect.

最好有音效。

So let me run this in cursor.

我现在用Cursor运行它。

I'll just paste it in.

我直接粘贴进去。

And I'm going to show an example that I've already generated just to save some time.

为了节省时间，我给你们看我已经生成好的示例。

So here is the beautiful castle that the model made.

这就是模型做出的漂亮城堡。

So it's just wild how, you know, from a concise prompt, the model has this great sense of aesthetics where it's made this like floating rock.

光凭简短的提示，模型展现了惊人的美学感觉，做出了悬浮岩石上的城堡。

made a 3D castle, and if you zoom in, you can see, like, tons of detail.

做了3D城堡，放大还能看到很多细节。

Like, these guards that are walking around, cannons firing.

比如巡逻的卫兵，还有开火的大炮。

Do you want to fire the cannons if you click this button?

想开炮吗？点这个按钮就能开炮。

Of course.

当然。

Who wouldn't want to?

谁不想呢？

There we go.

来吧。

So, you can fire the cannons.

你可以开炮了。

You can even chat with the characters.

你还能和角色对话。

So, we'll say hi to Captain Rowan.

我们来跟Rowan船长打招呼。

We have names.

角色都有名字。

Say hello to the merchant.

跟商人打招呼。

Merchants selling some stuff.

商人正在卖东西。

What's your favorite song?

“你最喜欢的歌是什么？”

A ballad of banners and dons.

“旗帜与贵族的民谣。”

Nice.

不错。

Give me some wisdom.

“给我一句智慧箴言。”

Curiosity is volatile.

“好奇心易燃易爆。”

Yeah, that makes sense.

很有道理。

Minigame.

小游戏。

Yeah.

来吧。

Do you guys want to try the minigame?

大家想试试小游戏吗？

Absolutely.

当然想。

Let's play the minigame.

我们来玩小游戏。

All right.

好。

So, you can fire at these balloons.

你可以射击这些气球。

Oh, wow.

哇。

All right.

好。

Oh, no, I'm not good at it.

不行，我玩得不好。

Hold on.

等等。

Maybe I can ask GPT-5 for some help with it.

或许我能让GPT-5帮帮我。

Oh, you hit one.

哦，你打中了一个。

Yes.

是的。

I got one.

我打中了一个。

We got a sound effect.

还有音效呢。

These are historically accurate balloons.

这些气球都是历史还原版。

Yes.

没错。

Working with GPT-5 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.

和GPT-5一起工作对我来说既有趣又意义深远，这是我第一次接触到真正具有创造力的模型。

And we're really excited to see how GPT-5 unlocks your creativity.

我们非常期待看到GPT-5如何激发你的创造力。