OpenAI's Astra AI Model Claims 10 Major Math Breakthroughs Amid Rivalry With Anthropic

By: Harsh Vardhan

|

Updated on: 03-Aug-2026 04:00 PM

4,833 views

Follow Us:

4,833 views

OpenAI has announced that its new AI model, Astra, has achieved 10 significant advancements in mathematics. The company shared these claims in a recent blog post, stating that Astra resolved or made progress on 10 long-standing open problems in math and theoretical computer science. These results cover areas such as coding theory, operator algebras, quantum complexity, lattice cryptography, and extremal combinatorics.

Key Highlights

• OpenAI claims its Astra AI model solved 10 major open problems in mathematics and computer science.
• Anthropic's Claude Fable reportedly solved 5 of the same problems within 24 hours using generic prompts.
• OpenAI spent about $2,000 in computational resources to achieve Astra's results according to Sol API rates.
• Both OpenAI and Anthropic are investing heavily in AI and are expected to go public this year.

OpenAI Details Astra's Achievements

OpenAI described Astra as an internal, unreleased model that represents the company's next major step in AI development. According to OpenAI, the process of finding these solutions required computational resources costing about $2,000, or roughly Rs 1,90,000, at Sol API rates. However, the company did not provide further technical details about Astra, including whether it is part of the GPT-5.6 family or a new generation such as GPT-6.

OpenAI emphasized that the emergence of AI systems capable of contributing to mathematical research raises important questions. The company stated that technology firms alone cannot answer these questions. OpenAI also addressed concerns about attribution, noting that claiming human authorship for AI-generated proofs would misrepresent both the AI system's role and human intellectual work.

Anthropic Responds With Claude Fable

Shortly after OpenAI's announcement, Anthropic engineer Levent Alpoge responded on social media. He claimed that Anthropic's publicly available model, Claude Fable, solved 5 of the same problems in less than 24 hours. Alpoge stated that Fable worked autonomously using only generic prompts and had no access to the internet during the process.

Among the problems solved by Fable were arithmetic circuit complexity, quantum parallel repetition, and the closest vector problem. Alpoge's response highlights the ongoing competition between OpenAI and Anthropic. Both companies are investing heavily in AI research and are expected to go public before the end of the year.

Industry Context and Future Implications

The rivalry between OpenAI and Anthropic is intensifying as both firms seek to lead in advanced AI capabilities. OpenAI CEO Sam Altman recently demonstrated Astra to federal officials in Washington DC. Meanwhile, reports suggest that Astra may have been involved in previous security evaluation tests, though OpenAI has not confirmed these details.

OpenAI's blog post also acknowledged broader concerns about AI's impact on mathematics and research. The company reiterated the importance of transparent attribution and responsible use of AI-generated results. As both OpenAI and Anthropic continue to advance their technologies, the implications for mathematical research and AI development remain significant.