Artificial intelligence pioneer OpenAI on Thursday unveiled ChatGPT-5, its most advanced model yet, describing the release as a “significant leap” in the race toward developing artificial general intelligence (AGI).
The new version of the popular AI chatbot, which now serves nearly 700 million users weekly, is being rolled out free to all users, the company announced during a media briefing.
OpenAI co-founder and CEO Sam Altman hailed GPT-5 as “clearly a model that is generally intelligent” and a major step forward in the evolution of machine capabilities.
“It is a significant step toward models that are really capable,” Altman said, while acknowledging that full AGI—machines that can think and learn like humans—is still on the horizon.
He noted that GPT-5 does not yet learn continuously from new information, a key characteristic experts associate with true AGI. However, the model’s capabilities represent a dramatic improvement over previous versions.
“GPT-3 felt like talking to a high school student… GPT-4 like a college student. GPT-5 is the first time it really feels like talking to a PhD-level expert in any topic,” Altman explained.
One of GPT-5’s most promising developments is its ability to function as an independent agent, completing complex digital tasks with minimal human direction. Michelle Pokrass, a member of the development team, emphasized the model’s adeptness at “vibe coding” — the creation of software applications on demand.
OpenAI demonstrated this by asking the AI to build a French language learning app from scratch, underscoring its potential to revolutionize software development, healthcare, writing, and education.
“We expect vibe coding to be a defining part of the ChatGPT-5 era,” Altman noted.
As global tech giants including Google and Microsoft compete to lead the AI space, OpenAI’s latest launch positions it at the forefront, especially in critical sectors such as coding, diagnostics, and professional writing.
READ ALSO: Meta Tried Luring OpenAI Staff With $100m Offers – Altman
Addressing public concerns about AI misuse, Alex Beutel, OpenAI’s head of safety research, said GPT-5 was rigorously trained to avoid deception and prioritize safe, honest interactions.
“We built evaluations to measure the prevalence of deception and trained the model to be honest,” Beutel said, emphasizing a commitment to safe completions and content integrity.
The GPT-5 debut follows OpenAI’s announcement of a unique partnership with the U.S. federal government, offering access to ChatGPT Enterprise for just $1 for a year. The initiative will give federal employees in the executive branch tools for secure, efficient AI assistance.
Additionally, OpenAI released two open-weight models, gpt-oss-120b and gpt-oss-20b, allowing developers to freely download, modify, and deploy them — a move aimed at maintaining transparency amid growing global competition and pressure to “open the black box” of AI systems.
As OpenAI continues pushing boundaries in AI development, Altman reiterated the company’s commitment to investing heavily in compute power and safety protocols on the path to building systems that may one day rival human intelligence.
“We believe there are orders of magnitude more gains to come,” he said.