When OpenAI rolled out GPT-5 on its platforms one day, Tesla CEO Elon Musk warned Microsoft CEO Satya Nadella that OpenAI could "eat" Microsoft alive.
Introducing the release, Nadella said, "Today, GPT-5 launches across our platforms, including Microsoft 365 Copilot, Copilot, GitHub Copilot, and Azure AI Foundry. It's the most powerful model to date from our partners at OpenAI, with dramatic new improvements in reasoning, coding, and chat, all trained on Azure."
Looking back on the rapid progress, Nadella noted that only two and a half years had passed since OpenAI CEO Sam Altman met up with him in Redmond to launch GPT-4 in Bing, describing the advancement as "incredible."
He went on, "The rate of progress is only going to speed up, and I have no idea what developers, businesses, and consumers will do with this next step."
In response to Musk's caution about OpenAI overtaking Microsoft, Nadella replied, "People have been attempting that for 50 years and that's the fun of it! You learn something new each day and innovate, partner, and compete. Looking forward to Grok 4 on Azure and awaiting Grok 5!"
At the same time, Cursor AI—a VS Code-based AI code editor—also affirmed incorporating GPT-5, referring to it as "the most intelligent coding model our team has ever tested" and stating it would be free "for the time being."
Musk, a supporter of the Grok AI platform, added, "Except that Grok 4 Heavy is still the most powerful AI."
Earlier, Sam Altman had spoken about his response after GPT-5 solved a tricky email problem he himself couldn't solve, calling it a "weird feeling" that made him feel "useless." While in a podcast, he equated the experience to the Manhattan Project, referring to physicist Robert Oppenheimer, and hinted at GPT-5 having "permanent effects" of such magnitude—albeit non-destructive in nature.
OpenAI launched ChatGPT-5 for free to everyone on Thursday, touting "major" improvements in AI abilities amidst increasing global competition. Altman called GPT-5 "clearly a model that is generally intelligent" and a "significant step" toward artificial general intelligence, but said it still is not constantly learning.
Contrasting the development over iterations, he explained that GPT-3 felt like a high school student, GPT-4 like a college student, and GPT-5 like a PhD-level specialist.
The new model performs well on independent "agent" tasks and "vibe coding," whereby it can create applications on demand while striving to be more reliable.
