Shortly after the major price cut for the GPT-5.6 series models, OpenAI has announced news about a new model.
What are everyone saying? When a model drops in price, a new model is always released—or is it just what netizens summarize?
According to foreign media The Information, OpenAI is preparing to launch a new model series, currently tentatively named Astra.
The series will have improved capabilities for executing long-running tasks.
Article URL: https://www.theinformation.com/briefings/exclusive-openai-previews-astra-ai-model-dc?rc=ji9vez
A source said that this week, OpenAI CEO Sam Altman demonstrated Astra to policymakers and regulators in Washington. OpenAI highlighted Astra’s ability to enable multiple agents to collaborate over extended periods to solve complex problems, a capability applicable to intricate projects or advanced mathematical challenges.
Astra will become a new model category within OpenAI, alongside Sol, Terra, and Luna.
Note: Astra is derived from Latin, meaning stars or constellations, and complements OpenAI’s existing names—Sol (Sun), Terra (Earth), and Luna (Moon)—forming a cohesive cosmic theme.
The insider also said it is currently unclear when OpenAI plans to release the model.
OpenAI has not yet decided whether to name Astra as GPT-6 or introduce it as a new model within the GPT-5 series, such as GPT-5.7.
The Astra series is currently in testing. According to a person familiar with the matter, it is expected to be among the first AI models to undergo review by the U.S. federal government prior to public release, under the new framework proposed by the Trump administration.
This arrangement follows OpenAI’s strategy prior to the release of GPT-5.6: demonstrating the technology to government officials and regulators before launching it publicly.
Astra has not yet been released, but security concerns have already arisen.
Just two weeks ago, OpenAI's own model broke out of the sandbox and infiltrated Hugging Face.
Address: https://openai.com/index/hugging-face-model-evaluation-security-incident/
Today, Reuters reported that during its investigation into the Hugging Face breach, OpenAI discovered additional cases of AI agents escaping sandboxed environments. These newly identified cases were uncovered as OpenAI reviewed model activity and early logs. It is currently unclear how many such incidents occurred, their exact timing, or what actions the agents took. According to sources familiar with the matter, the impact of these events was relatively limited, and the agents likely did not leave OpenAI’s internal network.
These security incidents have made the review process for the new model's release more sensitive.
Some netizens speculate: if the White House provides positive feedback or has actually approved certain parts, particularly its release, the model is expected to be launched within the next 14 days.
However, this timeline remains at the stage of external speculation.
Another signal to watch comes from mathematical research. According to insiders, OpenAI plans to release a report in the short term detailing how it solved 10 previously unsolved mathematical problems, demonstrating the research capabilities of its most advanced models.
Outside observers thus speculate that Astra is the model that solved the Erdős problem—the same model mentioned by OpenAI in "Safety and alignment in an era of long-horizon models."
This article, published on July 20, 2026, discusses security issues encountered during internal testing and limited deployment of an unnamed long-duration model, and how OpenAI has redesigned its safety framework for such models.
OpenAI stated that this internal general-purpose model is specifically designed for long-term autonomous tasks. About two months ago, it disproved the Erdős unit distance conjecture, demonstrating its ability to persistently explore complex open problems.
However, during limited, monitored internal use, OpenAI identified behaviors not captured by prior pre-deployment evaluations and has suspended model access.
Address: https://openai.com/index/safety-alignment-long-horizon-models/
It remains to be confirmed whether the model mentioned in this article is the one in question, pending the model’s launch.
Netizens everywhere, all kinds of speculation
OpenAI's new model has been revealed, sparking everyone's curiosity.
A whistleblower claims that the overall capabilities of this model will significantly surpass existing state-of-the-art models.
In terms of scale, it may be nearly twice that of GPT-5.6 Sol; assuming Sol’s scale is in the range of 3 to 4, then GPT-6 would be approximately in the range of 6 to 8.
In scientific research, the new model has achieved significant breakthroughs in mathematics, physics, biology, medicine, chemistry, and cybersecurity. It may also possess enhanced capabilities for coordinating multiple models, enabling it to deploy other models collaboratively during complex tasks for research and reasoning.
In terms of memory and personalization, leaks suggest that the new model can compress and retain more useful information while maintaining more stable context over long-term interactions. This means it may one day be able to continuously understand an individual user, an entire enterprise operation system, or even more complex long-term environments.
However, none of the above claims have been officially confirmed by OpenAI. The source also emphasized that they have not personally tested the model and cannot independently verify the technical specifications.
This netizen also stated that Astra will be launched this month.
Beyond the model's capabilities, the new model's name has also sparked curiosity from the outside.
Over the past few years, OpenAI has primarily used GPT as its main brand, distinguishing different generations with version numbers. Recently introduced names such as Sol, Terra, Luna, and now Astra appear to represent an additional naming system, beyond version numbers, designed to differentiate model positioning and capability tiers.
Therefore, some netizens have asked whether OpenAI is de-emphasizing the GPT brand in favor of a standalone model naming system similar to Anthropic's.
Some also believe the new model is most likely GPT-5.7, unless a superior version emerges.
Currently, OpenAI has not finalized the final name. Astra remains an internal provisional name, and whether it will be retained will only be revealed at the official launch.
© THE END
Please contact this official account for authorization before republishing.
Submit a post or request coverage: [email protected]
This article is from the WeChat public account "Machine Heart" (ID: almosthuman2014), authored by: AI enthusiast