OpenAI has introduced the name of its next-generation artificial intelligence family, Astra, while revealing that one of its internal models has achieved significant progress on some of mathematics’ most challenging unsolved problems.

According to the company, an unreleased Astra model either solved or made substantial progress on 10 long-standing open mathematical problems, highlighting what OpenAI describes as a major leap in AI reasoning capabilities.

The announcement comes as AI companies continue pushing beyond chatbots and coding assistants into advanced scientific research.

Tackling some of mathematics’ toughest challenges

OpenAI said the model worked across a broad range of mathematical disciplines, including high-dimensional geometry, coding theory, arithmetic circuit complexity, group theory, operator algebras, quantum complexity, lattice cryptography and extremal combinatorics.

Rather than simply generating answers, the AI produced complete mathematical arguments. Researchers later converted those arguments into formal academic manuscripts and verified each proof using Lean, a proof assistant widely used for mathematical verification.

The company also plans to publish the model’s reasoning process for every solution, giving researchers a closer look at how the AI arrived at its conclusions.

A step forward in AI reasoning

Although OpenAI has not disclosed technical specifications for Astra, it offered one interesting comparison. The company estimated that solving all ten problems would require computing resources roughly equivalent to $2,000 worth of GPT-5.6 Sol API usage.

While that figure is only an estimate, it suggests increasingly capable reasoning models may be able to tackle complex research problems without requiring prohibitively expensive computing resources.

The announcement reinforces OpenAI’s broader strategy of developing AI systems capable of working on long-running scientific and research tasks rather than simply responding to prompts.

Building on earlier research

This is not the first time OpenAI has claimed a mathematical breakthrough.

Earlier this year, the company revealed that another experimental model produced a disproof of the Erdős unit-distance conjecture, a famous mathematical problem that had remained unsolved for more than eight decades.

Taken together, these developments suggest AI is beginning to play a larger role in mathematical research, although researchers continue to debate how much credit machine-generated discoveries should receive.

Mathematicians urge caution

OpenAI’s latest announcement arrives at a time when many mathematicians are calling for greater transparency around AI-assisted research.

The Leiden Declaration on AI and Mathematics, published earlier this year, warned that AI-generated proofs raise questions about reliability, attribution, research integrity and proper disclosure.

Some researchers argue that while AI can become a valuable assistant, mathematical discoveries still require rigorous independent verification before they can be accepted by the wider academic community.

Responding to those concerns, OpenAI said it respects the ongoing discussion and believes research should clearly distinguish between the AI system that generated the mathematical ideas and the human researchers who formalised and verified them.

Expanding AI access for researchers

Alongside Astra, OpenAI recently announced ChatGPT for Academic Researchers, an initiative that will provide 100,000 scientists and mathematicians with free access to its latest AI models.

The programme is designed to help researchers explore difficult scientific and mathematical problems while encouraging responsible use of AI in academic work.

A glimpse into AI’s next chapter

While Astra is not yet publicly available, OpenAI’s announcement offers an early indication of where its research is heading.

The company is increasingly focusing on AI systems capable of contributing to scientific discovery rather than simply generating text or code. Whether Astra ultimately transforms mathematical research remains to be seen, but its reported achievements suggest advanced reasoning is becoming one of the next major frontiers in artificial intelligence.