HomeArticle

Countdown to the release of GPT-5? A full-network leak has emerged: Microsoft Copilot is preparing a major move, and GPT-5 is in the final sprint for launch.

新智元2025-08-01 10:00
GPT-5 is approaching its release. Multiple companies are preparing for integration, and the progress of AGI has drawn attention.

New revelations about GPT-5 are here! GPT-5-Alpha is already being internally tested by the Cursor team and can almost complete any task in one go. Perplexity has finished preparing for the GPT-5 release on its website. Microsoft engineers are also gearing up for GPT-5 and will soon launch it in Copilot. GPT-5 is really getting closer.

Every morning when I open my eyes, I can feel GPT-5 getting a little closer.

Just now, a whole bunch of new news about GPT-5 has emerged.

First of all, GPT-5-Alpha is being internally tested by the Cursor team. This model has an amazing performance and can almost complete any task in one attempt.

For example, the model successfully completed the following "Aquarium Game" challenge.

And just a few hours ago, someone discovered the GPT-5-Auto and GPT-5-Reasoning models in the macOS ChatGPT app.

Some sharp - eyed netizens noticed the word "reasoning" in the leaked information and speculated that this indicates that GPT-5 probably already has an o - series model.

Meanwhile, Perplexity has finished preparing for the GPT-5 release on its website. Once GPT-5 is released, Perplexity Pro users will be able to use it immediately.

Microsoft Prepares to Release GPT - 5 in AI Suite

Meanwhile, someone has discovered that Microsoft's Copilot Smart Mode will be powered by GPT-5.

Some people speculate that the router part of GPT-5 may be gradually being rolled out.

In short, Microsoft is preparing to release GPT-5 in its AI suite, including Copilot (consumer version), Microsoft 365 Copilot (enterprise/work version), and Azure (enterprise/API customers).

It is reported that the Copilot application on Windows 11 has an "Intelligent" mode that can switch between the reasoning/non - reasoning modes of GPT-5 based on the query.

Even the free version of Windows 11 Copilot will have an "Intelligent" mode based on GPT-5, so GPT-5 will not be limited to paid users.

Microsoft Engineers Gear Up for GPT - 5

Meanwhile, The Verge has just published an article stating that Microsoft is preparing to launch a new Copilot intelligent mode for GPT-5.

According to insiders at Microsoft, Microsoft is testing the intelligent mode of GPT-5 in both the consumer and business versions of Microsoft 365 Copilot, which is consistent with the previous revelations.

It is also said that in the consumer version, the AI in this mode can "think deeply or quickly according to the task", so users don't need to select a model.

This is also in line with what Altman previously revealed.

Altman said in February, "We hate model selectors as much as you do and hope to return to the magical unified intelligence."

Meanwhile, GPT-5 will include the o3 model instead of releasing it as a separate version.

The Verge reporter speculates that the early appearance of this intelligent mode in Copilot is because Microsoft engineers are preparing for the release of GPT-5.

In short, if OpenAI's preparations for GPT-5 go smoothly, the intelligent model of Copilot will soon be available to everyone.

Above is the latest wave of revelations about GPT-5 today.

Some netizens sighed that the development cycle of large models is now so fast that the marketing speed can't keep up with the release speed.

OpenAI Researcher: I Believe in AGI Again

With GPT-5 on the verge of arrival, at this moment, OpenAI researcher Alexey Guzey published an article titled "Why I Believe in AGI Again".

In this article, perhaps we can catch a glimpse of various signs of GPT-5.

The following is the general idea of the article.

Why Do I Believe in AGI Now?

First, I'm now convinced that ChatGPT can understand what it reads. Second, the reasoning model makes me believe that ChatGPT is creative. Third, ChatGPT can summarize text extremely well, which I think is a reliable standard for measuring intelligence.

Meanwhile, I don't believe in "general intelligence", so I think the concept of AGI doesn't make much sense.

Finally, AI products can now fund AGI research, which means AI has reached the early stage of a self - improvement cycle.

Therefore, many discussions about the timelines of "AGI" and "superintelligence" are outdated.

ChatGPT Can Understand What It Reads

For me, AGI is related to understanding. Does ChatGPT really understand, or is it just a Chinese Room stupidly mapping inputs to outputs?

I now think it really understands.

What really convinced me was a tweet. Someone mocked OpenAI's o1 model for not finding that the logical puzzle he proposed actually had a very simple solution.