这四个新AI模型如何改变你的工作方式

今天在《AI每日简报》中，我们来看看这周新上线的四个模型，会如何改变你的工作方式。

Today on the AI Daily Brief, how the count them four new models we got access to this week will change how you work.

《AI每日简报》是一档每天更新的播客和视频节目，聚焦AI领域最重要的新闻和讨论。

The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI.

好了，各位。

Alright, friends.

在正式开始之前，先说几件事。

Quick announcements before we dive in.

首先，感谢今天的赞助商：毕马威、Robots and Pencils、Blitzy 和 Airtable。

First of all, thank you to today's sponsors, KPMG, Robots and Pencils, Blitzy, and Airtable.

想收听无广告版本？请前往 patreon.com/aideallybrief，或者在 Apple 播客订阅我们。

To get an ad free version of the show, go to patreon.com/aideallybrief, or you can subscribe on Apple Podcasts.

如果你对赞助节目感兴趣，欢迎发邮件到 sponsors@aideallybrief.ai 告诉我们。

And if you wanna learn more about sponsoring the show, send us a note at sponsors@aideallybrief.ai.

欢迎回到《AI每日简报》。

Welcome back to the AI Daily Brief.

在人工智能给职场带来的诸多变化中，有一项几乎彻底打破了过去大家默认的共识——夏天热的时候，可以稍微放慢节奏。

Among the many changes that AI is bringing to the professional world, one of them is an almost total obliteration of the previously agreed upon idea that you could actually slow down a little bit over the warm summer months.

虽然不是每个白领都认为七月和八月就该休息、度假、补觉，但不可否认，这段时间确实会比平时安静不少。

While not every white collar professional would agree that July and August are a time for resting and vacations and catching up, it's pretty undeniable that it's a season where things quiet down a bit.

但在AI圈子里，情况完全不同，尤其是自从政府干预打乱了原有的节奏后，接下来恐怕会迎来一波接一波的新模型，正如你们待会儿会看到的，这些模型对我们的工作方式影响深远。

Except in AI Land, where almost especially now that the previous cadence was thrown off by the government's interference, we are in for what I believe will just be an absolute cavalcade of models, many of which as you'll see I think have fairly significant implications for how we work.

‘模型月’一开始就很热闹，Fable重新回归了——虽然严格来说它六月就发布了，但对我们大多数人来说，感觉上它更像是七月初才上线的。

The month of models got off to a big start with the return of Fable, which while yes, of course, technically was released in June, for all intents and purposes for most of us, it actually feels like an early July release.

而就在本周，我们又迎来了一大批新模型：OpenAI首次回应Fable的5.6 Sol，Grok推出的新成员Grok 4.5（这是他们和Cursor合作后的首个新品），Cognition发布的SWE-1.7，延续了Cursor在Composer 2.5中展现的趋势，最后，也是我们今天要重点聊的——GPT Live。

And yet this week, we added a whole new slate of models to the roster, including OpenAI's first answer to Fable, 5.6 Sol, a new entrant from Grok (their first since they hooked up with Cursor, called Grok 4.5), a new model from Cognition SWE-1.7, which continues a trend that we saw with Cursor in Composer 2.5 and finally the model that we're going to start with today, GPT Live.

你可能已经在社交媒体上看到过这个公告。

Now you might've seen this announcement floating around social media.

视频里是一群可爱又风趣的老奶奶，和ChatGPT的新版实时模型轻松聊天，目的是想展示这个新模型有多自然、多像真人对话。

It's a set of cute and charismatic grannies talking to ChatGPT's new live model in a way that's meant to represent just how much more natural and conversational the new model feels.

这当然不是今天节目的核心内容，但值得提一句：我觉得OpenAI这次的内容可能是他们近期最成功的宣传之一。

Now this is not at all the main point of the show, but it is worth noting as a side story that I think it may think that this content was some of the more effective we've seen from OpenAI.

Not Boring的帕基·麦科马克写道：OpenAI‘回归正常’的策略正在完美奏效。

Not Boring's Paki McCormack wrote, The OpenAI Just Be Normal strategy is working beautifully.

更妙的是，让这些女性显得聪明又优雅，明显掌控着对话节奏，甚至能打断、对聊天机器人说点不客气的话。

Bonus points for making the ladies look very smart and sophisticated and clearly putting them in control of the conversationinterruptingeven being rude with chat.

我觉得这个选择恰恰反映了这次发布的一个核心意图，那就是他们对消费者如何与AI互动的全新设想。

Now that particular choice I also think reflects one of the big underlying points of this announcement, which is an evolution in how they imagine consumers interacting with AI.

不过我们稍后再深入聊这个。

We'll come to that in just a minute though.

咱们先来谈谈这次发布的核心模型。

Let's talk about the actual model that was released first.

这个模型叫GPT Live，分两个版本：GPT Live 1 和 GPT Live Mini。

The model is called GPT Live, and it comes in two flavors, GPT Live one and GPT Live Mini.

它基于一种全双工架构，也就是说，OpenAI称它能同时听和说。

It is built on what they call a full duplex architecture, meaning OpenAI says that it can listen and speak at the same time.

这里重点在于，交互方式本身发生了重大变化。

And a big part of the emphasis here is in how the interaction model has changed.

早期的ChatGPT语音功能是基于一种叫级联语音系统的架构。

The earliest versions of ChatGPT voice were built on something called a cascaded voice system.

说白了，这个系统其实是三个模型串联在一起的。

Basically, that model was actually three models chained together.

用户说话后，先由语音识别模型转成文字，后台的大型语言模型生成回复，再由文字转语音模型还原成声音。

The user would speak, and then a speech to text model would transcribe that speech, an LLM in the background would produce a response, and then a text to speech model would convert it back into speech.

OpenAI自己也说，这套方案让我们第一次能和前沿AI模型对话，但复杂性带来了代价。

This approach OpenAI wrote enabled us to talk to frontier AI models for the first time, but the complexity came at a cost.

信息在模型之间传递时容易丢失，回复也慢得卡顿。

Information could be lost across models, and responses were slow and stilted.

接下来的版本，也就是集成在ChatGPT高级语音模式里的，是基于轮流对话的系统。

Now the next version, which was built into ChatGPT Advanced Voice Mode, was a turn based model system.

它在一个模型内部直接生成语音，不再需要经过语音转文字和文字转语音的步骤，延迟问题因此大幅改善，但依然是一问一答的轮流模式。

It generated audio within a single model as in it didn't have to go through that speech to text and text to speech translation, which made a big difference in terms of latency, but it still operated in a discrete turn based way.

我。

I.

模型必须等用户说完才能开始回应，这就带来了一系列问题。因为一旦依赖静音来判断说话结束，用户稍作停顿思考，或者背景有点噪音，系统就可能误以为对方说完了，结果插话插得特别突兀，对话显得生硬，一点都不自然。

E, the model had to wait for the user to stop speaking before it started, which created all sorts of challenges because when turn detection was based on silence, something like a brief pause in thought or background noise could be mistaken for the end of the user talking, which could lead to the model interrupting in weird ways, keeping things feeling stilted and not all that natural.

新模型做了两个重大的架构调整。

The new model makes two big architectural changes.

第一个是全双工架构，也就是说，GPT Live 是为持续互动而设计的。

The first is that full duplex architecture, which means that GPT Live, they say, is built for continuous interactions.

它不再是一条一条处理独立的对话，而是边听边说，实时处理输入的同时生成输出。

Instead of processing a sequence of separate messages they write, GPT Live continuously processes input while generating output.

据他们说，模型每秒能多次做出判断：该说话了，还是继续听，该暂停，该打断，还是该调用工具。

The model, they say, can therefore make interaction decisions many times per second, whether to speak, continue listening, pause, interrupt, or invoke a tool.

从他们展示的各种测试和视频里你都能听出来，这种互动方式跟人与人之间自然聊天的感觉，已经接近多了。

And as you can hear from all of the different tests and videos they shared, this makes for a conversational experience that's a lot closer to what you get when you're interacting in a normal human to human conversation.

第二个架构上的改变是，他们把 GPT Live 这个专注于持续对话的模型，跟那些更复杂的任务——比如推理、代理能力或者搜索——分开了。

The second architectural decision is that they separated GPT Live as a model, which has its focus on continuous interaction, from deeper types of work, including reasoning, agentic capabilities, or search.

也就是说，GPT Live 在对话进行的同时，可以随时调用另一个模型，比如 GPT 5.5，或者以后的 5.6，去处理搜索或推理这类任务，而对话本身完全不会被打断。

So basically, as the GPT Live model is doing that interaction, it can summon another model like GPT 5.5 or now going forward 5.6 to go take care of a task like searching or reasoning even as it keeps the conversation going.

几个月前，Thinking Machines Labs 推出过一个交互模型，做法跟这个很相似，你可能还记得。

Now you might remember a couple months ago when Thinking Machines Labs introduced their interaction model that took a pretty similar approach here.

实际上，莱利·布朗写过：这个来自 Thinking Machines 的图示，跟 GPT 语音版的发布很相关。

Riley Brown, in fact, wrote: This graphic from Thinking Machines seems relevant to the GPT voice release.

据我了解，两者确实非常接近。

They are, from my understanding, very similar.

实时语音模型内置一个交互模块，可以同时调用后台模型来使用工具。

Real time voice models have an interaction model that can run background models that use tools.

你跟它说话的时候，不用等它完成某个操作，它能自己在后台处理。

You don't need to wait for the voice model to finish doing something while talking to it, it can just do things in the background.

开发者丹尼尔·维拉说：我用这套架构做了一个项目。

Developer Daniel Vella wrote, I built a project with this architecture.

我的项目不是一个备用的后台大模型，而是一个以语音为主的微型代理，会把任务交给 Hermes 或 OpenCode 这样的其他代理，通过 ACP 来协调。

Instead of a secondary background LLM, my project is a voice first mini agent that delegates tasks to other agents like Hermes or OpenCode through ACP.

你跟它聊天的时候，哪怕后台在跑重任务，对话也能一直继续。

You can continuously talk with the agent while heavy task occurs in background.

比普通语音助手舒服多了。

Much more pleasant experience than the usual voice agents.

某种程度上，你看到的这种模式，其实和当前先进AI系统中越来越常见的架构很像——多个模型同时运行，由一个协调模型启动其他代理或子代理，在后台完成特定任务，最后整合成一个连贯的回应。

So in some ways, what you're seeing here is a voice equivalent version of an increasingly common architectural pattern in general with advanced AI systems that involve multiple models at the same time and the ability for an orchestrator model to initiate agents or sub agents to go do some specific set of tasks that work in the background before organizing it all into a coherent single response.

从这些演示中，马上就能看出几个有意思的地方。

So there are a couple interesting things that come up right away from the demos.

很多人第一时间想到的，就是这种技术在多语言交流场景下的巨大潜力。

One of the first places that people went was just how powerful this was going to be for anything involving multiple language interactions.

实时、无停顿的交互意味着，模型可以边说边翻译，根本不用等说话人说完再处理。

Live, turnless operation means, for example, that the model can translate each phrase as it's spoken, not having to wait for a break in the conversation.

这样一来，模型的表现就更像真人翻译，而不是死板的机器。

This means the model can function much more like a human translator and less like a wooden computer.

想想看，你有没有见过那种演讲翻译，译者不是等演讲者说完才开口，而是在对方刚开口几秒后就开始同步翻译。

Think about when you've seen a translator for a speech where the translation happens just a few seconds after the person being translated starts speaking, rather than waiting for that speaker to stop and pause at any given point.

人们也很快意识到——这里不玩文字游戏——这玩意儿简直是语言学习的绝佳工具。

Now people also grok very quickly, no pun intended, that this seemed like a really excellent tool for language learning.

Click Health的西蒙·史密斯说，没错，这玩意儿学语言真的特别棒。

Click Health's Simon Smith says that yes indeed for learning this does rock.

他提到，我试过用它练习西班牙语。

He wrote, I tried practicing Spanish with it.

它能给我英文单词让我翻译，还能精准点评我的发音。

It can give me words in English to translate and properly critique my pronunciation.

他还觉得，这工具的用途远不止语言学习，如果有个通用的闪卡工具，咱们能用它来学各种东西。

He also though saw that this was probably valuable for more than just language learning, noting that if there was a generic flashcard tool, we could use this for all sorts of learning.

他觉得，像闪卡、测验、类似黑板的涂鸦板这些功能，用在这里简直太完美了。

He writes flashcards, quizzes, Blackboard like Sketchpad, and so forth would just be incredible here.

这东西完全可以当个超级老师。

This thing could be an exceptional teacher.

不过得提一句，虽然现在还没有专门的闪卡功能，但新版ChatGPT已经能展示天气、股票、体育、日程等信息的可视化卡片了。

Now one thing that is worth noting is that although there is not a flashcard feature specifically, the new model in ChatGPT does now have the ability to show visual cards for things like weather, stocks, sports, schedules, and more.

这说明，加个闪卡功能其实并不难。

This suggests that the jump to something like flashcards wouldn't be all that difficult.

不过我觉得，这些闪卡其实属于另一个故事的延伸，这个故事始于那些老人，而在我看来，这清楚地表明了这款产品真正瞄准的使用场景，就是我们常说的‘Siri式’需求。

And yet the flashcards, I think, are another part of the story that started with the grannies, which that is, to me, this is very clearly the set of use cases that this is really focused on, are what we might call the Siri use cases.

这就是那些普通用户——不是用ChatGPT来工作，而是用来帮自己处理日常事务的人——所依赖的基础个人助手功能。

This is the basic personal assistant stuff that they assume a lot of regular people who aren't using ChatGPT for work using ChatGPT for.

再回头看看那则奶奶的广告，他们让新模型做的很多事情，其实正是过去几年里，很多人在社交媒体上抱怨 Siri 做不到的那些事。

When you look at the Grandma's ad again, a lot of the things that they ask the new model to do are the things that you've seen for the last several years people screeching about on social media because it seems like stuff that Siri should be able to do but can't.

他们实际上是在用这些典型失败场景来测试系统，看看它会不会崩溃，而结果是，系统表现得很稳定。

They're basically stress testing it along archetypal failure modes that cause user frustration, and it appears to keep functioning properly.

我觉得，这种以语音为主的交互方式，可能比现在主流的打字方式，更接近未来人们与AI互动的形态。

And I think in some ways that this sort of interaction mode, voice first, might be closer to the future of how people interact with AI than the current paradigm of typing.

萨姆·阿尔特曼就说过类似的话，他写道：‘GPT Live 今天已在 ChatGPT 中上线。’

Sam Altman basically said as much, writing: GPT Live launches today in ChatGPT.

用起来感觉既神奇又真实。

It feels magical and real.

我一直以来都更喜欢打字跟AI交流，而不是说话。

I have always preferred typing to talking to an AI.

不过我觉得这种情况马上就要变了。

Now I think that's going to shift.

有意思的是，虽然这对普通消费者来说非常合理，但更值得探讨的是，这对职场用户的工作互动方式又会带来什么影响。

What's interesting though, is that while I think this makes a ton of sense for consumers, it's a more interesting question to ask how this might change the interaction pattern for work based users as well.

经常听我节目的朋友应该知道，我最常推荐的一个提升AI使用效率的方法，就是改用语音模式。

Now, regular listeners will know that one of my most common tips to getting more out of AI is to switch to a voice paradigm.

不是改用这些语音模型，而是像WhisperFlow这样的工具，让你能随口说，而不是打字输入。

Not to switch to these voice models, but to use something like WhisperFlow to be able to ramble speaker inputs rather than typing them.

原因其实很简单。

The reason for this is pretty simple.

AI的表现好坏，很大程度上取决于你给它的上下文信息。

AI tends to live or die on the basis of the context you give it.

不管你打字多快，说话的速度几乎肯定更快。

And no matter how fast you type, you almost certainly talk faster.

打字还会强迫你把想法结构化，这虽然有时有用，但未必总能帮上你真正想用AI达成的目标。

Typing also forces you into structure in a way that, while it can be useful, it might not always actually be helpful to what you're trying to achieve with AI.

简而言之，你说话的时候，

In short, when you speak, I.

嗯，

E.

随口说一说，就能给AI提供更多上下文。

Ramble, you can provide the AI a lot more context.

现在，我哪怕坐在电脑前，绝大多数和AI的互动也都用语音完成了。

And at this point, the vast, vast majority of my interactions with AI, even when I'm sitting at my desktop computer, are done via voice.

但这不是那种一问一答的语音模式。

But it's not this sort of back and forth voice mode.

我只是把语音当输入方式。

It's just about voice as my input.

我根本不在乎AI要不要回话，像我们和Super Intelligent一起做的那个采访系统，我就发现，等AI慢慢念出回答反而特别烦人，还拖慢节奏。

Not only do I not care about having the AI speak back to me, in many applications including the one that we built with Super Intelligent that interviews people I've sometimes found that waiting for the AI to speak back can be incredibly annoying and slow.

毕竟，你的大脑理解一句话的速度，远快于听AI用那种慢吞吞的语速念出来。

Basically, your brain's ability to read a sentence and absorb it is a lot faster than having to listen to the AI talk at some mediated pace.

所以我对这个新模型有个有趣的问题，它会不会开始影响人们的工作方式，如果会的话，会怎么影响？

So an interesting question I have about this new model is how it might start to interact, if at all, with people's work behaviors.

虽然现在还早，但在这方面已经有一些有意思的反馈了。

And while it's early, there is some interesting feedback on that front.

早期测试者达里亚·阿努特·马兹说：自从语音版和ChatGPT最初发布以来，我很少用它。

Early tester Daria Anut Maz wrote: Since the original release of Voice and Chat GPT, I rarely used it.

但这个新版本彻底改变了我的使用习惯。

But this new version completely changed that.

我现在成了它的常客。

It has made me a frequent user.

它的语音听起来自然多了，流畅、聪明，和它聊天真的特别舒服。

The voice is so much more natural, smooth, intelligent, and really so enjoyable to talk with.

我还试了用我的母语土耳其语，包括一些地方方言，效果出奇得好。

I also tried it in my mother tongue, Turkish, including a few local dialects, and it does a phenomenal job.

实时翻译确实令人印象深刻，但更让人惊讶的是，它能捕捉到语言的细微差别、节奏和地道的感觉。

The simultaneous translation is impressive, but what is even more remarkable is that it captures the subtleties, rhythm, and local feeling of the language.

好吧，现在有个原本持怀疑态度的人开始转变了，这跟莱利·布朗的经历也很像。

Okay, so we have a skeptic who's starting to be converted, which was also Riley Brown's experience.

他写道：这个结果真出乎意料。

He wrote, This one was unexpected.

我以前对语音模型根本提不起兴趣，直到试了这个。

I was not excited for voice models until I tried this one.

它太真实了，简直有点吓人。

It's very realistic, almost spooky.

最厉害的是，这个模型能快速上网搜索。

The coolest part, the model can search the Internet fast.

如果你用过 GPT 实时版，就应该知道，现在的语音模型在使用工具方面已经越来越熟练了。

And if you've spent any time with GPT Real Time too, you know these voice models are getting very good at using tools.

我猜用不了多久，你就能通过跟这个语音模型聊天来处理邮件、查业务数据、安排会议，几乎任何事都能搞定。

Soon, I'm guessing, you'll be able to do your email, check on business data, schedule meetings, and basically anything just by chatting with this voice model.

实现像贾维斯那样的技术，现在基本已经到位了。

The tech is basically already there for Jarvis.

现在只剩下生成式界面这一层了。

The only thing left is a generative UI layer.

所以这里我们看到的不仅是莱利从怀疑者转变了态度，他甚至开始意识到，这项技术未来可能改变他的工作方式，哪怕现在还没完全实现。

Okay, so we have here Riley not only being a skeptic converted, but starting to see how it might start to become something that changes how he works, even if it's not there yet.

西蒙·史密斯又写道：这让我第一次觉得，AI 真正像一个能干的数字助手，就像贾维斯那样。

Simon Smith again wrote, This felt like the closest I've seen an AI come to being a capable human like digital assistant, Jarvis, her, what have you.

结合我过去两周使用 Claude Tag 的体验，我感觉我们正处在一个转折点——AI 越来越像同事，而不再只是工具。

On the heels of my experience the past two weeks with Claude Tag, it feels like we're in a transition period where AI's become more like colleagues and less like tools.

伊森·马莱克也表达了类似的观点。

Ethan Malek made a similar point.

GPT 语音功能非常出色，也更接近科幻作品里和 AI 对话的体验，部分原因就在于它更聪明了。

GPT Voice is really good and much closer to the science fiction experience of talking to AI in part because it is much smarter.

再加上 Claude Tag，这暗示着人机协作方式正在出现新的形态。

That plus Claude Tag suggests emerging new modes of working with AI.

你可能还记得，Claude Tag 是那个嵌入在 Slack 里的 Claude 版本，它能在多个用户频道间共享上下文，而不仅仅局限于单个用户的独立实例。

Now Claude tag, you'll remember, is the version of Claude that sits inside your Slack and has shared context across user channels, I e, isn't just limited to a single user's instance of Claude.

马克斯·温巴赫写道，想象一下GPT Live集成在Codex里，还配有线程管理工具。

Max Weinbach wrote, imagine GPT Live in Codex with the thread management tools.

你只要和它聊天，它就能分配任务、阅读内容，并管理项目和聊天中的所有工作。

You just chat with it, and it can delegate, read, and manage all the work going on in projects and chats.

我觉得这迟早会实现，而更进一步，就是用语音控制你整个电脑。

I imagine this will be a thing, and downstream of that is just controlling your entire computer with voice.

普林斯有个精彩的比喻：2025年，AI是个玩具。

Prince had a great analogy writing, twenty twenty five, AI is a toy.

2026年，AI成了住在瓶子里的精灵。

Twenty twenty six, AI is a genie that lives in a bottle.

只要你清楚瓶子在哪，也知道怎么许愿，你的愿望就会实现。

If you know where to find the bottle and how to phrase your wish, then your wish shall be fulfilled.

2027年，精灵已经逃出瓶子，和你朝夕相伴。

2027, the genie has escaped the bottle and lives alongside you.

它能从上下文推断你的需求，在你开口前就主动完成。

It infers your wishes from context and fulfills them before you ask.

最重要的技能是实时引导这个精灵。

The most important skill is Real Time Genie Steering.

所以，一切都不同了吗？

So, is everything different?

事情真的全变了吗？

Did everything change?

别急着下结论，盖尔·魏纳写道：我刚试了GPT语音功能，声音本身太棒了。

Not so fast, says Gail Weiner, who wrote: I just tried GPT Live Voice, and the voice itself is gorgeous.

语速、温度感和自然度都真的令人印象深刻。

The pacing, the warmth, the naturalness is genuinely impressive.

听起来就像个跟你一起思考特别舒服的人。

It sounds like someone you'd enjoy thinking with.

但声音本身不是产品，真正重要的是思考能力。

But the voice isn't the product, the thinking is.

再动听的声音，如果包裹的是肤浅的推理，也不过是徒有其表。

A beautiful voice wrapped around shallow reasoning is just a pretty face with no depth.

看着挺舒服，但帮不上你真正做事。

It's pleasant to look at, but doesn't help you build.

如果你只是想快速找答案，那确实不错。

If you're looking for quick answers to questions, it's lovely.

但如果你想深入讨论、一起思考，那就真不太行。

If you want to brainstorm and work, it's not great at all.

Husq 发的这段视频进一步印证了这一点：虽然语音模型变聪明了，但在真正的智力表现上，还不一定算得上顶尖。

Reinforcing that point is this video from Husq, who showed that while the voice model may have gotten smarter, it still isn't necessarily state of the art when it comes to raw intelligence.

ChatGPT 刚刚更新了语音功能，据说现在聪明多了，也自然多了。

Chat GPT just had a big voice update, and this is supposed to be way smarter now and way more natural.

那咱们来实际测试一下。

So let's put that to the test.

我已经把它调到最智能的模式了。

I have it set to the highest smart setting.

你好。

Hey.

你能告诉我数字17里有多少个字母e吗？

Can you tell me how many e's are in the number 17?

有两个e。

Two e's.

一个在seven里，一个在teen里。

One in seven and one in teen.

你真聪明啊。

And you're really smart.

对吧？

Right?

是的。

Yeah.

17里确实有两个e。

Two e's in 17.

你确定吗？

And you're sure about that?

当然了。

Absolutely.

S-E-V-E-N-T-E-E-N。

S e VENTE EN.

没错，就是这样。

That is too.

没错，就是这样。

I that's right.

完全正确。

Exactly.

这明显是个梗，OpenAI那边也挺淡定的，阿迪·阿莱蒂还说：兄弟，你们连一天都没给我们留。

Now this is obviously meant to be a bit of a meme, and the OpenAI folks took it in stride with Adi Aletti saying, bro, you didn't even give us a day.

开玩笑的，我们超爱你的视频。

JK, we love your videos.

但我觉得这里关键在于，当你对比GPT Realtime 2的API和GPT Live 1，还有GPT-5这类推理模型时，它们不只是模型层级的差别，更是交互体验的差别。

But I think that the important thing here to note is that when you look at the difference between GPT Realtime two, the API, and GPT Live one, and the GPT-five class reasoning models, these are not just model tiers, they're interaction tiers.

Realtime 2 是专为开发者打造的更智能的实时语音模型。

Realtime two is the smarter real time voice model for developers.

GPT Live 1 是 ChatGPT 的语音交互体验。

GPT Live one is the ChatGPT voice experience.

它能提供更自然、可随时打断、流畅连贯的界面，后台还能调用 GPT 5.5 的能力。

It's a more natural, interruptible, always flowing interface that can call on GPT 5.5 behind the scenes.

但如果你像 Gale 那样，想做那种深度策略性辩论——这也是我最常用的一个场景——真正的衡量标准还是底层的前沿推理模型，而不是语音层本身。

But if you are looking to do something like it sounds like Gale was, and something that's one of my main use cases, which is extended strategic debate, the benchmark is still going to be the underlying frontier reasoning model, not the voice layer itself.

换句话说，语音模型其实是在做权衡。

Voice models, in other words, are making a trade off.

它们得一边听、一边判断什么时候插话、怎么自然回应、避免尴尬的沉默，还要把延迟压到最低。

They have to listen, decide when to interrupt, respond naturally, avoid awkward silence, and keep latency low.

这跟静下心来深入思考用户提出的问题，完全是两个不同的优化目标。

That's a very different optimization target from sit with the problem the user is suggesting and reason deeply.

但即便如此，这依然是一个信号：整体来看，哪怕不是这个具体模型，交互方式已经开始转变了。

And yet still, this is one more indication that overall, even if it's not this model exactly, the interaction pattern is starting to shift.

Sitebringer提到：GPT Live标志着AI不再像软件，而开始像一个活生生的认知存在。

Sitebringer wrote: GPT Live points at the moment AI stops feeling like software and starts feeling like a live cognitive presence.

文字让AI保持距离，语音则消除了这种距离。

Text keeps AI at a distance, voice collapses that distance.

一旦模型能自然地打断、做出情感回应、跟上节奏、察觉犹豫、听出不确定性，并跟上人类的思维节奏，互动就从‘使用工具’转向了‘建立关系’。

Once the model can interrupt naturally, respond emotionally, follow rhythm, catch hesitation, hear uncertainty, and keep pace with human thought, the interaction starts moving from tool use into relationship use.

当然，这在AI的陪伴属性方面会带来深远影响，我们以后可能会聊到，但我认为，单纯把这看作是消费者层面的变化是错误的。

Now obviously there are implications here for the sort of companionship dimension of AI, which we'll probably discuss at some point in the future, but I think it would be a mistake to solely view this as a consumer shift.

相反，我认为这其实是更大趋势的一部分——随着我们把越来越多工作交由AI处理，我们的工作重心转向协调与管理，语音可能会成为我们进行这种协调越来越重要的方式。

Instead, I think it is part and parcel of a larger shift where as we delegate more and more, and more of our work becomes coordination and management, voice might be an increasingly important way that we do that coordination.

当前最重要的AI问题之一，不是谁在用AI，而是谁用得好。

One of the most important AI questions right now isn't who's using AI, it's who's using it well.

普华永道和德克萨斯大学奥斯汀分校刚分析了140万次真实职场中的AI互动，发现了一个令人意外的现象：最有影响力的人并不是更擅长写提示词的人，而是把AI当作思维伙伴来使用。

KPMG and the University of Texas at Austin just analyzed 1,400,000 real workplace AI interactions and found something surprising: the highest impact users aren't better prompt engineers they treat AI like a reasoning partner.

他们会提出问题、引导思考、反复迭代，并推动得出更好的答案。

They frame problems, guide thinking, iterate, and push for better answers.

那好消息是什么？

And the good news?

这些行为是可以大规模培养的。

These behaviors are teachable at scale.

如果你正想从使用AI提升到真正掌握AI能力，KPMG关于高级AI协作的研究非常值得一看。

If you're trying to move from AI access to real capability, KPMG's research on sophisticated AI collaboration is worth your time.

了解更多请访问 kpmg.com/us/sophisticated。

Learn more at kpmg.com/us/sophisticated.

网址是 kpmg.com/us/sophisticated。

That's kpmg.com/us/sophisticated.

我总看到一些公司到处押注，云平台、模型、框架全试一遍，或者付钱给系统集成商做永远做不完的试点项目。

One thing I keep seeing in companies hedging across every cloud, every model, every framework or paying a GSI for a pilot that never ends.

真正有成果的团队已经选好了方向，正在快速推进。

The team's actually shipping, they've picked a lane and they move fast.

这正是我喜欢今天赞助商 Robots and Pencils 的原因之一。

That's one of the reasons I like today's sponsor Robots and Pencils.

他们全面投入了 AWS。

They've gone all in on AWS.

他们是 AWS 高级层级合作伙伴，能在 45 天内交付生产级的 AI 助手。

They're an advanced tier and AWS pattern partner and they ship production AI coworkers in forty five days.

这让他们在 AI 助手领域做出了我见过的一些最有趣的应用。

That's led to them doing some of the more interesting work I've seen on AI coworkers.

我说的不是聊天机器人，而是真正嵌入企业架构、能执行实际任务的智能代理系统。

And by that I'm not talking about chatbots, I'm talking about actual agentic systems that sit inside a business architecture and do real work.

如果你是企业领导者，想把真实成果落地，或者你是 AWS 销售，想把客户从感兴趣变成真正部署，这种专注就特别关键。

That kind of focus matters if you're an enterprise leader trying to get something real into production, or an AWS rep trying to move a customer from interested to deployed.

前往 robotsandpencils.com 预约 AI 专题简报。

Request an AI briefing at robotsandpencils.com.

和 Robots and Pencils 聊一次，你就会明白了。

One conversation with Robots and Pencils and you'll know.

想把企业软件开发速度提升五倍吗？

Want to accelerate enterprise software development velocity by 5x?

你需要Blitzy，这是唯一专为大型企业代码库打造的自主软件开发平台。

You need Blitzy, the only autonomous software development platform built for enterprise codebases.

你的工程师只需定义项目、新功能、重构，或者全新构建的任务。

Your engineers define the project, a new feature, refactor, or greenfield build.

Blitzy的智能代理会先全面扫描并梳理你的整个代码库，然后生成一套量身定制的执行方案，供团队审阅和确认。

Blitzy agents first ingest and map your entire codebase, then the platform generates a bespoke agent action plan for your team to review and approve.

一旦获得批准，Blitzy就会自动开始生成数十万行经过验证、端到端测试过的代码。

Once approved, Blitzy gets to work autonomously generating hundreds of thousands of lines of validated, end to end tested code.

一次运行就能完成超过百分之八十的工作量。

More than 80% of the work completed in a single run.

Blitzy不是在写代码，而是在以计算的速度开发软件。

Blitzy is not generating code, it's developing software at the speed of compute.

你的工程师只需审阅、优化，然后发布即可。

Your engineers review, refine, and ship.

这就是财富500强企业如何把原本需要数月的项目压缩到一个冲刺周期内完成的方式。

This is how Fortune 500 companies are compressing multi month projects into a single sprint.

将工程效率提升五倍。

Accelerating engineering velocity by 5x.

前往 blitzy.com 亲自体验 Blitzy。

Experience Blitzy firsthand at blitzy.com.

就是 blitzy.com。

That's blitzy.com.

本集《AI每日简报》由 Hyperagent 赞助播出，您可以在这里管理团队协作的智能代理群组。

This episode of the AI Daily Brief is brought to you by Hyperagent, where you run fleets of agents your team can manage together.

新用户可获得 1000 美元的推理算力。

New users get $1,000 in inference.

别再依赖本地代理或等着在电脑上手动触发的聊天流程了。

Forget local agents and chat workflows waiting on your laptop to be prompted.

Hyperagent 将智能代理长期部署在云端，无缝对接您团队正在使用的各类工具，持续执行实际任务。

Hyperagent deploys always on agents in the cloud, doing real work across the tools your team already uses.

市场团队的代理能自动将竞争对手的动态转化为落地页面。

Marketing's agent turns competitor moves into landing pages.

销售团队的智能代理负责完善潜在客户信息、撰写邮件，并更新客户关系管理系统。

Sales's agent enriches leads, drafts emails, and updates the CRM.

运营代理则跟进文书工作，追踪预算使用情况。

Ops agent chases the paperwork and tracks the budget.

每个代理都能访问共享的上下文信息，并严格遵守你设定的范围和审批规则。

Every agent has access to shared context and follows your rules about scope and approvals.

是时候引入那些像同事一样可靠的智能代理了。

It's time you add agents that feel like teammates.

立即前往 Hyperagent，由 Airtable 团队打造，为你配备专属代理。

Hire yours at Hyperagent built by the team at Airtable.

前往 hyperagent.com/aideallybrief 领取你的 1000 美元推理额度。

Claim your $1,000 in inference at hyperagent.com/aideallybrief.

这个

The

接下来我想聊一聊的是 Grok 4.5。

next new model I want to discuss is Grok 4.5.

说实话，直到最近——特别是Cursor被收购的消息公布之前，很多人都以为埃隆可能会逐渐退出模型层，转而专注于算力层。

And candidly, up until very recently, specifically the announcement of the Cursor acquisition, there have been a fair number of people who have thought that we might be seeing Elon moving away from owning the model layer to owning the compute layer.

当然，他后来确实改变了想法，开始重新评估自己的竞争优势。通过向Anthropic、谷歌，甚至其他一些公司开放Colossus和Colossus 2，埃隆明显决定也要在算力霸主这个层面参与竞争。

Certainly, at some point, he started to think differently about his competitive advantage By giving both Anthropic and Google and even some other companies access to Colossus and Colossus two, Elon was clearly deciding that he was willing to play at that compute czar type of level as well.

而SpaceXAI上市后股价大涨、在公开市场取得成功，其中一个结果就是他们手头有了更多资金，而这些资金就成了强大的杠杆。

And yet, one of the outcomes of SpaceXAI going public, and seeing such a big pop, and having success in the public markets, is that they have a lot more capital to play with, and that capital becomes leverage.

Cursor收购案的顺利完成，就是一个绝佳的例子。

The completion of the Cursor acquisition was a great case in point.

Grok 4.5正是SpaceX AI与Cursor全新合作后的首个成果。

And Grok four point five is the first output of the new collaboration between SpaceX AI and Cursor.

SpaceX AI称这是他们首个专门为编程和智能体训练的模型，特别强调工程类应用场景。

SpaceX AI called this their first model trained specifically for coding and agents, and really emphasizes those engineering type of use cases.

他们说，Grok 4.5是为真实世界的工程开发打造的。

Grok 4.5, they wrote, was built for real world engineering.

它在处理大型代码库时表现优异，能应对跨越多个仓库、数百项技能和多种工具的长期任务。

It excels in large codebases and handles long running tasks that span multiple repositories, hundreds of skills, and a variety of tools.

Cursor方面在推特上表示，这是他们迄今为止最强大的模型，也是首个不仅面向软件工程设计的模型。

Cursor, for their part, tweeted, It's our most powerful model yet, and the first we've built for more than software engineering.

对于担心Cursor已转向不同方向的人，他们的官方推特账号特别说明，Grok 4.5和Composer属于不同规模的模型权重，Composer 2.5不仅会继续提供，未来还会推出更多同级别的新模型。

Now for those worried that Cursor is now headed in a different direction, the main Twitter account made sure to point out that Grok 4.5 and Composer are two different model weight classes, and that not only will Composer 2.5 continue to remain offered, but they will be releasing new models of that size going forward.

你可能还记得，Cursor 2.5是基于Kimi架构，并用他们独有的数据进行微调，从而以更低的成本实现高性能的模型。

Cursor 2.5, you'll remember, is their model that was built on a Kimi base and post trained using their unique data to get high performance at a lower cost.

在各项基准测试中，Grok 4.5在Terminal Bench 2.1、SWE-bench Pro和Deep Suite 1.0上的表现基本与Opus 4.8和GPT-4.5持平。

When it comes to the benchmarks, Grok 4.5 is more or less a match for Opus four point eight and point five across Terminal Bench 2.1, SWE-bench Pro, and Deep Suite one point zero.

在人工分析指数上，Grok 4.5总体排名第四，落后于Fable 5、Opus 4.8和GPT-5 0.5。

On the Artificial Analysis Index, Grok 4.5 places fourth overall behind Fable five, OPUS 4.8, and GPT-five 0.5.

它在智能体类知识工作和编程方面表现尤为出色，在GDP Valve榜单中也位列第四。

It has particularly strong performance in agentic knowledge work and coding, and also ranks fourth in GDP Valve.

但更重要的是，我们越来越清楚，这些模型如今的竞争不再只看纯粹的智能水平，还越来越看重效率。

Importantly, though, we continue to see that these models are no longer simply competing on the question of raw intelligence but also on questions of efficiency.

在人工分析测试中，Grok 4.5是近前沿模型中成本效益最高的一个。

On artificial analysis tests, Grok 4.5 was by far the most cost efficient run of the near frontier models.

在整个基准测试中，每个任务的成本仅为31美分，而Opus 4.8要1.80美元，Fable 5则要2.75美元。

It cost 31¢ per task across the overall index benchmark as compared to $1.80 for Opus four point eight and two seventy five for Fable five.

在某些对比中，它的成本略低于GLM 5.2和Kimi K 2.6，但稍高于Haiku 4.5和Nemotron 3 Ultra。

For some comparisons, it was slightly cheaper than GLM 5.2 and Kimi K 2.6, and slightly more expensive than Haiku 4.5 and Nemotron three Ultra.

它的单任务成本大约是GPT-5.5的三分之一、Opus 4.8的五分之一，以及Fable 5的九分之一。

Its cost per task was about a third of GPT-fivefive, a fifth of Opus 4.8, and one ninth of the cost of Fable five.

单从这些基准数据来看，说它以Haiku级别的成本实现了接近Opus的性能，完全说得通。

Effectively, based on the benchmarks alone, it is very realistic to say that the model delivers near Opus performance for haiku level costs.

与此同时，Grok Build也相当值得关注。

Meanwhile, Grok Build seems like a big deal as well.

Build搭配Grok 4.5在AI编程代理指数中排名第三，超过了Opus 4.8搭配Claude Code，仅次于GPT-5.0.5搭配Codex和Fable 5搭配Claude Code。

The combination of Build and Grok 4.5 was the third best harness and model pairing on AI's Coding Agent Index, ahead of Opus 4.8 in Claude Code, and just behind GPT-five 0.5 in Codex and Fable five in Claude Code.

Grok 4.5在自动化基准测试中也达到了顶尖水平，这是AI公司自研的评测体系，用于测试模型在模拟的Excel、Gmail和Slack等SaaS应用环境中调用工具完成真实工作流的能力。

Grok 4.5 is also fully state of the art on Automation Bench, which is AI's proprietary benchmark that tests a model's ability to use tools to complete a range of real world SaaS workflows in simulated app environments like Excel, Gmail, and Slack.

简单说，这是一个基于真实场景的智能体式计算机操作评测。

Essentially, it's an agentic, computer use style benchmark based in real world examples.

它不仅以51.4%的得分位居第一，超过了Fable five的48.6%，而且每项任务的成本只有0.34美元，远低于Fable的1.35美元。

And not only did it get the top score at 51.4% compared to, for example, Fable five's 48.6, it did that at $0.34 per task compared to $1.35 for Fable.

它使用的token数量不到Opus 4.8的一半，在使用工具调用时的交互轮次也比其他前沿模型更少。

It used less than half the tokens of Opus four point eight's run, and fewer turns while using tool calls, than other Frontier models.

但我们都清楚，基准测试有时会误导人。

But as we know, benchmarks can be deceiving.

今年四月，YouTuber兼AI创业者Theo曾写道：我真心觉得xAI可能会迎来一次惊人的逆转。

Back in April, YouTuber and AI entrepreneur Theo wrote, I legitimately believe xAI might have a crazy comeback.

在测试了Grok 4.5之后，他转发了自己的帖子，并说：被我猜中了。

And after testing Grok 4.5, he retweeted himself and said, Called it.

埃隆·马斯克后来也转发了Theo制作的关于Grok 4.5的视频，视频里他对这个模型给予了高度评价。

Elon Musk would actually later on retweet the video that Theo ended up producing about Grok 4.5, where he had a lot of positive things to say about the model.

在让模型同时处理多个代码修改和拉取请求后，Theo说：这种来回互动、多目标的复杂任务，大多数模型如果被拆得太细、上下文堆得太满，根本做不好。

After asking it to do multiple code based changes and PRs at the same time, Theo said, This is the kind of back and forth and complex multi target tasks that you could make most models do if you try to break them up really carefully and not try to bloat the context.

我对4.5的使用体验完全没有任何问题。

I had no ergonomic issues with 4.5 at all.

用起来真的很舒服。

It was actually really pleasant to work with.

它找到所有Fable发现的内容了吗？

Did it find everything Fable found?

没有。

No.

它的代码有GPT-5.6那么周全吗？

Was its code as thorough as GPT-5.6?

没有。

No.

它能不能在处理复杂繁重的任务时和我顺畅互动，同时又快又便宜？

Was it able to go back and forth with me on really big, heavy tasks and be pleasant to use while also being very fast and cheap?

是的。

Yeah.

在很多方面，Theo认为这个模型可以作为Opus 4.8的一个不错替代品，而且在不少地方还超过了GLM 5.2。

In a lot of ways, I see this model, Theo concludes, as a good alternative to something like Opus 4.8, and it kind of outclasses GLM 5.2.

我现在对这个模型已经没什么兴趣了，除了它是开源权重之外。

I have no interest in that model anymore aside from it being open weights.

但Grok 4.5意外地是个不错的默认代码模型。

But Grok 4.5 is a weirdly good default code model.

我觉得某种程度上，Grok 4.5是第一个有意识地理解自己在当前复杂模型架构时代中位置的模型。

And I think in some ways, what you can view Grok 4.5 as as the first model that self consciously understands its relationship to the state of the art in the era of more complex model architectures.

埃隆·马斯克向来不吝赞美，他曾发推说：公平来讲，Fable确实比Grok 4.5强，但大多数任务根本不需要Fable那种级别的能力。

Elon Musk, who is not one for underselling, even tweeted at one point, In fairness, Fable is definitely better than Grok 4.5, but most tasks don't require Fable level capability.

事实上，如果你看看早期使用者的反馈，他们兴奋的不是4.5能取代OpenAI或Anthropic的顶级模型，而是把它当作实现型代理使用，而让Fable或GPT-5.6来负责统筹协调。

And indeed, if you look across the early adopters, what they're excited about is not 4.5 replacing the top OpenAI or Anthropic model, but using it as the model for one of the implementation agents when you have Fable or GPT-5.6 acting as the orchestrator.

这恰好发生在非常关键的时刻。

Now this comes at an extremely important moment.

还记得昨天整个节目都在讨论，如果中国决定不再支持本国公司继续推动前沿开源模型，会带来什么后果和独特的市场机会吗？

Remember, yesterday's entire episode was about what happens and the unique market opportunities that arise if China decides to back off allowing their companies to continue to push state of the art open weight models.

对一些人来说，只要试过4.5，这个问题就不再重要了——他们根本就会直接选它，而不是去用GLM 5.2或者KimiK这类模型。

For some playing around with 4.5, that question ceases to matter as they're just going to default to that instead of one of those GLM 5.2 or KimiK models anyway.

Stonk Daddy提到，很多人还没真正意识到Grok 4.5发布意味着什么。

Stonk Daddy wrote, not sure people fully grasp what the launch of Grok 4.5 means.

企业已经开始收紧token的使用预算。

Enterprises are already clamping down on token spend.

中国开源模型在性价比上表现强劲，也逐渐被接受，但很多机构还是因为数据主权、合规、安全和地缘政治风险而避而远之。

Chinese open source models offer strong pricing performance and have gained traction, but many orgs still avoid them over data sovereignty, compliance, security, and geopolitical risks.

Grok 4.5的性能超越了中国开源模型，价格却接近它们，而且没有‘中国开源’的标签——它是西方打造的，对企业用户更友好。

Grok four point delivers better than Chinese open source performance at near Chinese open source cost without the stigma of being Chinese open source, Western built, and enterprise friendly.

它马上就要把其他模型的市场份额全抢光了。

They are about to suck the oxygen out of the room.

正如Stong指出的，对了，顺便说一句，Cursor早就进入企业了。

And as Stong points out, Oh, and surprise, Cursor is already in Enterprises.

这个‘特洛伊木马’已经帮Grok顺利进了门。

That Trojan horse already got Grok through the door.

现在我们又回到了新模型层出不穷的时代，这周我们还迎来了Cognition的SWE-1.7。

Now because we are back to being absolutely spoiled with new models, we also this week got Cognition SWE-1.7.

这跟 Cursor 的 Composer 模型更接近了。

Now this is closer to Cursor's Composer model.

和 Composer 一样，它基于 Kimi 的架构（这次是 Kimi K2.7）。

Like Composer, it's built on a Kimi base (in this case, Kimi K2.7).

他们提到，在标准基准测试中，它的表现略优于 GLM 5.2 和 Composer 2.5，几乎接近前沿模型，但完成同样任务的成本只有前沿模型的一半到三分之一。

They point out that it's slightly better than GLM 5.2 and Composer 2.5 on the standard benchmarks, and almost as good as Frontier models, but costs only around half to a third as much as those Frontier models for the same tasks.

目前还没多少人有机会实际体验 SWE 1.7，所以要判断它在真实场景中的表现还为时过早。

Now, not as many people have had a chance to play around with SWE 1.7 yet, so it's hard to get too many first impressions of its actual performance in the real world.

不过，Cognition 团队在谈论它的时候，有两点我觉得挺有意思。

There were two things that I thought were interesting, though, about the way that the Cognition folks were talking about it.

第一点是它对研究现状意味着什么。

First is the implications for the state of research.

Cognition 的创始人 Silas Alberdi 写道：最近，业界的关注点已经从强化学习转向了预训练和中间训练。

Cognition founder Silas Alberdi wrote: Recently, the industry chatter moved from reinforcement learning to pre training and mid training.

强化学习的回报会逐渐递减，因为它受限于预训练质量的天花板。

RL is supposed to get diminishing returns because it hits a ceiling that depends on the quality of your pre train.

这当然没错，但真正的上限到底有多高，目前还是个未知数。

While that is certainly true, it's an open question how high the ceiling actually is.

我们想把这一点推向极限。

We wanted to push this to the limit.

KimiKade 2.7 已经经过多次微调，一开始看起来几乎没什么提升空间了。

For Kimikade 2.7, which has already been heavily post trained multiple times, it initially seemed like there wasn't much to squeeze.

虽然遇到了快速熵崩溃和其他不少问题，但我们把这些当作挑战，反过来优化了算法和数据。

We dealt with fast entropy collapse and many other issues, however, we used this as a challenge to tune our algorithm and data.

突然有一天，它开始奏效了，SWE-1.7 的表现提升幅度远远超出了我们的预期。

At some point, it suddenly started working, and the model SWE-1.7 magically improved by much more than we expected.

看来上限还没真正触到，所以我们很期待还能再往前推进多少。

It also seems like the ceiling isn't obviously hit yet, so we're excited to see how much further we can push it.

Cognition 的 Jeff Wang 表示：我们在研究上取得了不少重大突破，比如多区域训练、自压缩技术、效率提升，还有整体行为上与基础模型 KimiKade 2.7 的显著不同。

Cognition's Jeff Wang wrote: We've made some awesome breakthroughs in research, such as multi region training, self compaction, and efficiency, as well as changing the overall behavior compared to Qimikaze 2.7, the base model.

Ben Dixon 指出，这其实反映了一种更普遍的趋势。

And Ben Dixon points out that this is a broader pattern.

他提到，应用层公司利用对用户体验数据的独特访问，训练出专门执行特定任务的模型，性能能媲美封闭式前沿AI模型，但速度更快、成本更低。

Application layer companies, he writes, are using their unique access to UX data to fine tune specialized models that perform specific tasks on par with closed frontier AI models at a higher speed and at a fraction of the price.

Cognition的SWE-1.7是第二个展现出良好效果的应用层模型。

Cognition SWE-1.7 is the second application layer model that is showing promising results.

Cursor的Composer系列也遵循同样的模式。

Cursor's Composer series follow the same pattern.

封闭式前沿AI模型适合用于探索。

Closed frontier AI models are good for exploration.

开源大模型才是实现规模化和生产落地的核心引擎。

Open LLMs are the engine for scale and production.

这种趋势我们越来越常见，也一直密切关注。

Now this is a pattern we're seeing more and more, and that we've been paying attention closely to.

另外我想提一下SWE-1.7的运行速度。

Now the one other thing that I wanted to flag around SWE-1.7 is its speed.

JP Zinaware说：刚试了SWE-1.7 Lightning版，简直快得离谱。

JP Zinaware writes, Just tried SWE-1.7 lightning and holy crap it's fast.

我眼睛都快跟不上了。

My eyes can't keep up.

真正让这不只是个新奇玩意儿的原因在于，它再次改变了人机交互的模式。

Now what makes this more than just a novelty is that this actually has implications once again for the interaction patterns.

去年，Sean Wang（也就是 Swix）画过一张图，讲的是 AI 编程面临的一个难题：要么你处理的任务需要频繁互动，得一直盯着；要么任务太复杂，干脆全扔给系统，自己先走开。但中间那段就特别尴尬——延迟太高，没法实时响应，又不够长到能安心离开，整个人特别别扭。

Last year, Sean Wang, aka Swix, had a great graph of one of the challenges for AI enabled coding where things worked great if you were either a, dealing with tasks that needed enough interaction that you were there and engaged or b) on the other end of the spectrum, dealing with tasks that were complex enough that you could just delegate them entirely and go away for a while, but that in the middle, there were real challenges because there was too much latency to be actually real time, but not enough to go away for long, making it really uncomfortable.

SWE-1.7 正好卡在中间这个地带，但它靠速度几乎重新定义了这些问题。

SWE-1.7 seems to be playing in that space, but almost challenging the problems based on its speed.

Cognition 的 Nader Davitt 写道：用这种方式开发，开创了一种全新的开发者体验——过去你得走开等结果的任务，现在还没来得及走神，它就干完了。

Cognition's Nader Davitt wrote: Building this way is a new category of DX, a task that used to justify walking away finishes before you've mentally moved on.

现在出现了一种奇怪的中间状态。

There's a weird middle mode now.

技术上还是异步的，但快到你根本不用走，就坐着看它完成。

It's technically async but fast enough that you just watch.

我觉得这个趋势特别值得观察，它其实也触及了我们今天要聊的最后一个模型——GPT-5.6——的核心议题。

This is something that I think is going to be super interesting to watch, and actually gets at one part of the conversation of the final model that we're going to discuss today, which is GPT-5.6.

关于GPT-5.6的早期评价陆续出来了，我们现在得到的不只是‘还不错’这种笼统的说法。

The early reviews of GPT-5.6 continue to come in, and we're starting to get more details than just it was good or not.

尤其是，我们开始更清楚地看到它在哪些方面表现突出，和Fable五代相比如何，以及我们打算怎么把它用在更复杂的新型模型架构里。

Specifically, we're starting to get a lot more about where it excels and how it compares to Fable five, and how we're likely to use it as part of our new, more complex model architectures.

IAI的Clairvaux提到：GPT-5.6明显强于其他模型的地方包括：浏览网页、像人一样写作、与人有效沟通、交付真正可用的产品而不只是代码，还有视频剪辑、前端设计和高效的工作循环。

How IAI's Clairvaux writes: Things GPT-5.6 Sol is significantly better at than other models: Browser use, writing like a human, communicating to a human, shipping usable product not just code, video editing front end design and productive loops.

Every公司的Dan Schipper是AI领域最严谨的文本测试者之一，他也认为，用他的话说，GPT-5.6比Fable写得好多了。

Dan Schipper from Every, who are some of the most conscientious writing testers in AI, also agrees that GPT-5.6 is, in his words, a much better writer than Fable.

Dan补充说：它能一次生成出以前所有模型都搞不定的营销邮件。

Dan continues: It consistently one shots marketing emails that every previous model would fail at.

Fable太啰嗦，还容易陷入自己编一套语言说话。

Fable is too verbose and liable to fall into using sentences in its own private language.

律师Prins发现，5.6在跟他律师工作相关的内容上表现非常出色。

Lawyer Prins found that 5.6 was very good at work that related to his job as a lawyer.

他写道：只要相关的法律依据都能在网上公开获取，它就能完全替代任何级别的助理律师，专门负责法律调研。

He wrote: It can replace an associate of any level in the specific task of legal research, provided that the entirety of relevant legal authorities are publicly available online.

这个说法虽然范围很窄，但这类法律检索确实是我律师工作中非常关键的一部分。

This is a very narrow claim, but this kind of legal research is a very important part of my work as a lawyer.

现在，Prinz之所以更倾向用OpenAI的模型而不是Anthropic的模型来做这类法律工作，是因为他有一个关键测试标准，就是搜索能力。

Now, part of why Prinz prefers OpenAI models to anthropic models for this type of legal work is that one of his key tests is around search.

他会测试模型在海量信息中精准定位关键内容的能力——这在法律领域尤其重要，因为那里需要极其精确的信息。

He tests a model's needle in the haystack search capabilities, which are obviously very important when you're dealing with the legal field which needs extremely precise information.

至于整体对比，大家特别喜欢用比喻来形容。

Now, in terms of overall comparisons, people were really plumbing for analogies.

Dan Schipper又写道：GPT-5.6就像一辆保时捷。

Dan Chipper again writes: GPT-5.6 is like a Porsche.

Fable则像曲速引擎。

Fable is like a warp drive.

他说，在内部测试了大约一个月后，Every团队一致认为，GPT-5.6在日常知识工作和编程中，综合表现出了最强的性能、速度和效率。

He said that after testing 5.6 internally for about a month, that the team at Every decided that it was the best combination of power, speed, and performance for your day to day knowledge work and coding.

Fable是完全不同的类型，Dan写道。

Fable is a different beast, Dan writes.

如果要横跨整个星系，就用 Fable。

If you need to get across the galaxy, use Fable.

如果只是日常通勤，用 5.6 这样最适合当前任务的工具就最合适。

If you need to get around town using the best available tool for the job, use 5.6.

事实上，Dan 写道：他们失去对 5.6 的访问权限后，他感觉就像回到了石器时代。

In fact, Dan wrote: When they lost access to 5.6 he said he felt like it was back in the stone age.

他团队里的另一位成员 Austin Tedesco 说，从 5.6 回退到 5.5，感觉就像用一个比平时重两倍的篮球投篮。

Another member of his team, Austin Tedesco, said that going back to 5.5 after having access to 5.6 felt like, quote, trying to shoot a basketball that's twice as heavy as the one I'm used to using.

再深入一点，Every 团队发现了一些值得注意的地方，我觉得你们在测试 5.6 时也该留意这些。

Going into some more details, there were a few things that Every found, which I think are things that you should look out for as you start to test 5.6.

他们提到的第一点是，速度会改变模型的表现。

The first, they said, is that speed changes the model.

Sol 是他们目前最信赖、最常使用的速度最快的模型。

Sol, they wrote, is the fastest model we trust as a daily driver.

每次修改只要几分钟，这样更容易丢弃不理想的草稿，换个方向重新尝试，而且问题还新鲜地留在脑子里，思路不会断。

A revision costs minutes, which makes it easier to discard a weak draft, try a different direction, and keep moving while the problem is still in your head.