- Published on
OpenAI is reportedly building Astra, a model family designed to work on problems for hours or days
- Authors
- Name
- DeepHorus
OpenAI is reportedly developing a new AI model family called Astra, designed for long-running tasks by coordinating multiple agents. CEO Sam Altman demoed the system to politicians and regulators in Washington this week. The project aims to push AI toward continuous reasoning over hours or days. The Information reported the details, citing three people familiar with the plans.
Astra would form a new model class alongside OpenAI's existing Sol, Terra, and Luna families. It may ship as GPT-6 or as a variant like GPT-5.7; no release date is set. The system is already being tested.
OpenAI stressed Astra's ability to coordinate multiple agents over extended periods for hard problems, such as complex projects and advanced math. OpenAI also plans to publish a report demonstrating how its most advanced AI solved ten previously unsolved math problems. The goal is to show what current models can already do.
Astra is expected to be the first model tested under a new U.S. regulatory framework. The Trump administration's plan would require AI models to be submitted to the federal government before public release. The administration aims to finalize the framework by the end of this week. This means Astra could face an unprecedented approval process.
A key challenge is whether Astra can avoid compounding errors during long workflows. It must correct itself when a process drifts off course as context grows. This remains a major weakness in today's agentic systems. Multi-agent setups can also underperform on tightly linked tasks, such as planning, because coordination overhead may cancel out gains.
The rumors align with earlier statements from OpenAI leadership. Chief Scientist Jakub Pachocki said last summer that OpenAI wants AI systems that can work on a problem for hours or days. Late last year, OpenAI raised the question of systems that could solve tasks requiring centuries of human work. These comments indicate a strategic shift toward long-horizon AI.
By March 2028, OpenAI wants a fully autonomous AI researcher. As early as September, it plans an AI system with research-intern-level skills to speed up human scientists. Astra could be that system. Pachocki also said these systems will need far more compute, reflected in OpenAI's long-term infrastructure plans.
Whether OpenAI's revenue grows fast enough to fund the buildout remains an open question. Astra represents a major step toward continuous, agentic AI, but success depends on solving reliability issues. The project also raises regulatory questions as the U.S. government moves to review advanced models. If Astra delivers, it could redefine AI's role in research and complex problem-solving.
