I reported last week that OpenAI is planning to launch GPT-5 in early August, as part of an effort to simplify and combine its large language models. Just a day after my report, references to GPT-5 were spotted inside Microsoft’s Copilot web app, hinting at a new smart mode.
Sources familiar with Microsoft’s AI plans tell me that the company is currently testing this new smart mode for Copilot across both the consumer version and the commercial Microsoft 365 Copilot. In the consumer version, the mode is described as offering an AI that “thinks deeply or quickly based on the task,” so you don’t have to pick different models.
The employee-only version of Microsoft 365 Copilot has a similar smart mode, which allows Copilot to “use the most relevant model for your request to give you better results,” I’ve learned. None of the internal versions of Copilot mention GPT-5 yet, and the model suggests it’s still using GPT-4 when answering queries. But this UI still points toward a GPT-5 mode for Copilot.
OpenAI CEO Sam Altman revealed earlier this year that the AI lab is working toward improving ChatGPT’s model picker. “We hate the model picker as much as you do and want to return to magic unified intelligence,” Altman said in February. In the same post on X, Altman also revealed that GPT-5 will include its o3 model instead of shipping it as a standalone version.
This idea of a “magic” model picker has also been used by Microsoft internally. I understand that some parts of Microsoft 365 Copilot have been showing a magic mode in recent weeks that’s designed to function the same as the smart mode. It’s likely that magic mode is simply a codename for smart mode, and this magic version still doesn’t mention GPT-5. That’s not all that unusual, though, as Microsoft doesn’t usually list the OpenAI model it’s using for its various Copilot modes.
I suspect that this smart mode is showing up in Copilot early because Microsoft engineers are preparing for the release of GPT-5. OpenAI had previously targeted an earlier release of GPT-5, and Microsoft has typically followed up with its own implementation of OpenAI’s models in Copilot quickly after they’re released. Microsoft’s AI-powered version of Bing was using OpenAI’s GPT-4 model for six weeks before OpenAI officially announced GPT-4. Microsoft also quickly launched OpenAI’s o1 reasoning model as part of a Copilot overhaul last year, before making it free to use several months later. Microsoft even made the Sora video generator free to use, months after OpenAI released its paid version.
Microsoft declined to comment on the GPT-5 references in Copilot, and the company isn’t commenting on its new smart mode, either. If all goes well with OpenAI’s final GPT-5 preparations, then I expect we’ll see Copilot’s smart mode show up for everyone very soon.
I’m always keen to hear from readers, so please drop a comment here, or you can reach me at notepad@theverge.com if you want to discuss anything else. If you’ve heard about any of Microsoft’s secret projects, you can reach me via email at notepad@theverge.com or speak to me confidentially on the Signal messaging app, where I’m tomwarren.01. I’m also tomwarren on Telegram, if you’d prefer to chat there.
Thanks for subscribing to Notepad.
