Before the end of this year, Sam Altman announces that OpenAI will have an internal system which he will designate as general artificial intelligence. However, the leader admits that the comp
Before the end of this year, Sam Altman announces that OpenAI will have an internal system which he will designate as general artificial intelligence. However, the leader admits that the company has not yet crossed this stage. This projection is mainly based on Astra, a new family of models capable of performing research tasks over several days. However, no autonomous test can certify that OpenAI is really on the threshold of AGI.
According to statements by Sam Altman, OpenAI should reach general artificial intelligence. The leader attests that the company is “not quite there yet”, but it will have by the end of this year an internal system that he would qualify as AGI.
He emphasized:
We have not quite yet reached AGI, but by the end of the year, OpenAI will have internally a system that I would describe as such.
The word choice provides an essential reservation. Altman neither confirms a public launch nor the official crossing of a universal technological threshold. He shares his own judgment of a model still in development and evaluated within OpenAI.
For Mark Chen, the company’s research director, OpenAI is currently at “80% of the way” towards AGI. However, no public protocol allows reproducing this calculation. Such a percentage thus reveals the judgment of the laboratory’s leaders, not the result of a standardized scientific indicator.
As for Greg Brockman, co-founder and president of OpenAI, he adopts an equally optimistic stance. For him, observers would consider in two years that AGI emerged during the current period. These statements demonstrate that management no longer treats this technology as a distant goal.
The chosen definition remains essential. AGI is then described in OpenAI’s official charter as a set of independent systems capable of surpassing humans in the majority of activities with economic value. This approach therefore grants a central position to professional performance, without requiring that the machine replicate all human faculties.
The entire trust of OpenAI rests on Astra, its upcoming family of models. During a presentation to clients, 16 agents succeeded in dividing among themselves a mathematics problem of a research level. Furthermore, they managed to coordinate their work before providing a demonstration proposal.
Astra would also have the ability to use various software on a computer and conduct missions over a long period. These agents would work like virtual colleagues without waiting for new instructions after each stage.
According to Jakub Pachocki, OpenAI’s chief scientist, the company has already reached its internal goal related to automating the work of a beginner researcher. Thanks to an experimental concept, Astra would have the capabilities to write the necessary code within OpenAI systems, start the experiment, then present its results. The model could also perform some tasks that would occupy a researcher for nearly a week.
This potential would help accelerate the development of next-generation models. An artificial intelligence capable of conducting experiments might contribute to improving its successor. Each advancement should thus shorten the next cycle’s duration, a process usually called recursive improvement.
However, the company remains more cautious in its official communication. In a plan released in June, OpenAI specified that a significant part of its research work would simultaneously be conducted by humans and AI systems by March 2028. Achieving a first step this year does not yet constitute automation of all its scientific activities.
The feats mentioned through on-chain data have not yet been subject to a detailed technical publication. The company has presented no complete results, no comparison with human researchers, nor any evaluation led by an external structure. Astra remains an internal system whose abilities are fundamentally described by its creators.
Start your crypto adventure safely with BybitThis link uses an affiliate program.There is a difference between the system that Sam Altman might personally qualify as AGI and an official company announcement. The topic has no universally accepted definition. That is why a model might excel in programming or research while remaining less proficient in new situations, physical work, or decisions requiring a deep understanding of the real world.
Partnerships between OpenAI and Microsoft include another phase. Since their revision in October 2025, an official AGI communication by the company must undergo a thorough review by a team of independent experts according to Microsoft’s statement. Altman’s personal opinions would therefore not suffice to trigger the contractual obligations related to this technology.
Such verification has financial consequences. The partnership especially provides that profit sharing between the two entities continues until the panel confirms AGI attainment. Also, Microsoft retains certain rights over OpenAI’s models and products until 2032, including systems designed after this stage.
This projection comes as OpenAI has slowed down some tasks to consolidate its safety measures. Astra’s release would depend on new evaluations, without a concise public timeline. A delay in commercialization would not prevent the company from testing a more advanced system internally. However, it should limit the possibility to verify its capabilities.
Ultimately, it should be emphasized that four elements will allow judging this forecast by the end of this year: the release of Astra’s evaluations, tests conducted by external researchers, a possible official declaration from OpenAI, and the decision of the independent panel planned with Microsoft. The AGI announced by Sam Altman will fundamentally remain an internal qualification in their absence.
