OpenAI is putting a near-term marker on one of the most contested goals in technology: artificial general intelligence. According to a TIME report, company leaders believe AGI is close, but the meaning of that claim depends on a definition that remains disputed across the field.
Sam Altman says OpenAI is "not quite yet" at AGI, while also saying the company will have an internal system he would call AGI by the end of 2026. Chief Research Officer Mark Chen puts OpenAI at "80% of the way" to AGI, and co-founder Greg Brockman believes this period may later be seen as the moment AGI emerged.
Astra sits at the center of the AGI argument
The model family drawing the most attention is Astra, OpenAI's upcoming system. OpenAI has described it as an automated AI research intern, and Chief Scientist Jakub Pachocki told TIME that Astra already reaches that internal benchmark.
In practical terms, the reported benchmark is about research execution. Given an experimental idea, Astra can write code inside OpenAI's codebase, run the experiment, and return results. Given a research paper, it can do work that would have taken a human researcher about a week.
That framing matters because it shifts the AGI discussion away from chat alone. The focus is not just whether a model can answer difficult questions, but whether it can take a complex research task, perform the steps, and report back in a useful way.
Astra is also tied to what OpenAI calls persistent agents. These are virtual coworkers that can work on tasks over longer periods rather than responding only in short, isolated interactions. That capability is part of why Altman sees Astra as more than another incremental model release.
"I expect this will be the first model where the model actually invents new things in a way that matters,"
Altman said that during a customer preview, according to TIME. He added: "That's a very AGI-like thing."
The definition of AGI is doing a lot of work
The key question is not only what Astra can do, but what counts as AGI. The term has no precise definition in the source report, which makes OpenAI's claim partly a matter of interpretation.
OpenAI defines AGI as "highly autonomous systems that outperform humans at most economically valuable work." That is an economics-driven definition. It emphasizes autonomy and performance in work that has economic value, rather than requiring a broader philosophical or scientific test of intelligence.
Other views are more demanding. The source notes that some argue AGI requires a robust understanding of the world, and that large language models are only one part of the puzzle. Under that view, a system that handles many valuable tasks may still fall short if it cannot truly understand, generalize, or discover in a deeper sense.
This disagreement explains why OpenAI's confidence can be both significant and controversial. If AGI means a highly autonomous system that beats humans at most economically valuable work, Astra-like agents may look close. If AGI means broad world understanding and reliable discovery across domains, the evidence is less settled.
Discovery and self-improvement remain disputed
The biggest implication of Astra is the possibility that AI could help improve AI. The source describes the underlying wager as recursive self-improvement, where artificial intelligence contributes to making itself better.
Opinions are split. Some researchers believe that kind of capability is still far off. Others see early signs that it may already be happening.
The debate also reaches into the limits of language models. A central question is whether AI systems built purely on language models can truly make new discoveries and generalize. The source frames that ability as a basic requirement for AGI, but also makes clear that there is no settled answer.
For readers, this is the practical takeaway: OpenAI is not merely claiming faster chat or better coding. Its leaders are pointing toward systems that can act, persist, run experiments, and possibly generate new knowledge. Whether that deserves the label AGI depends on which standard is applied.
OpenAI is also preparing products and hardware
The AGI push is not only a research story. OpenAI is also working on devices and product changes that could bring more agentic AI into everyday use.
Altman told TIME that OpenAI is developing a "small handful" of devices: "something that belongs on a table," something for your pocket, and something you wear. The first is expected to be a small, puck-shaped device that senses its surroundings and talks to its owner through voice mode. TIME reports it is expected in early 2027.
Altman also said OpenAI will "definitely" build humanoid robots. The source does not describe timing or technical details for those robots, but the statement shows that OpenAI's ambitions extend beyond software interfaces.
On the product side, OpenAI is combining Codex, its coding agent, with ChatGPT through an internal project called "The Merge." That work produced ChatGPT Work, which is intended to put agentic AI in front of a mass audience.
The difference is task handling. Instead of only answering questions, the system is meant to act with more persistence. Thibault Sottiaux, the product lead running the project, told TIME that OpenAI is "close" to shipping something with "persistence and always-on execution." He also said: "It's definitely going to feel like a new thing to people."
The business problem is part of the technology story
OpenAI's ambitions come with a more ordinary challenge: paying for them. The source says the company's answer is tighter product focus and new revenue experiments.
One area is advertising inside ChatGPT. According to TIME, early tests have performed well enough that leadership is increasing the volume. CFO Sarah Friar said ad revenue could help cover costs for the 92 percent of ChatGPT consumer users who do not pay for a subscription.
OpenAI is also testing "Sponsored Agents," where clicking an ad moves users into a brand-hosted AI experience. That points to a business model in which agentic systems are not just tools for work, but also new surfaces for commerce and brand interaction.
The result is a high-stakes convergence of research, products, hardware, and revenue strategy. OpenAI's leaders say AGI is within reach by their definition. The rest of the debate turns on whether systems like Astra can truly invent, generalize, and sustain useful work at a level that justifies the name.