OpenAI's Astra Solves Ten Unsolved Maths Problems for $2,000

30-Second Summary

On 1 August 2026, OpenAI announced that its next major model, internally codenamed Astra, had produced verified solutions to ten open problems in mathematics that have resisted expert effort for between ten and thirty years. The solutions span group theory, high-dimensional geometry, coding theory, quantum complexity, lattice cryptography, and extremal combinatorics. Each solution was formally verified using Lean, a proof assistant that requires every logical step to be machine-checkable. OpenAI estimates the total compute cost at approximately $2,000 in tokens. Astra is not yet commercially available, but the announcement establishes that AI systems can now perform at expert research level in technical domains, at a cost that is a small fraction of what equivalent human expert work would require.

At a Glance

• Topic: AI Strategy
• Company: OpenAI
• Date: 1 August 2026
• Announcement: OpenAI disclosed that its next internal model, Astra, solved ten long-standing open problems in mathematics
• What Changed: AI has produced formally verified solutions to decades-old expert problems at a compute cost of approximately $2,000
• Why It Matters: Expert-level cognitive reasoning is no longer cost-prohibitive, signalling a fundamental shift in which knowledge tasks AI can take on
• Who Should Care: Business owners in knowledge-intensive sectors including legal, financial, consulting, research, and any organisation that pays for expert analytical work

Key Facts

• Company: OpenAI
• Announcement Date: 1 August 2026
• What Changed: Astra produced formally verified solutions to ten long-standing open problems in mathematics, verified via the Lean proof assistant
• Cost: Approximately $2,000 in compute tokens
• Who It Affects: Knowledge workers, professional services businesses, and any organisation that relies on expert-level analytical or research capacity
• Primary Source: OpenAI (via multiple outlet reporting including Forbes, The Next Web, The Decoder, DataCamp)

What Happened

OpenAI published findings on 1 August 2026 showing that its next internal model, referred to by the codename Astra, had resolved ten mathematical problems that have remained open for between ten and thirty years. The problems span six technical fields: group theory, high-dimensional geometry, coding theory, quantum complexity, lattice cryptography, and extremal combinatorics.

Among the confirmed results, Astra constructed an explicit example of a non-sofic group, resolving a question that has been open since Mikhail Gromov introduced the concept of soficity in 1999. It also disproved the Connes rigidity conjecture on von Neumann algebras, proved Ehrhart's volume conjecture, and resolved three problems from Paul Erdos's published catalogue of unsolved problems. Each solution was verified in Lean, a formal proof assistant that requires every logical step to be spelled out in machine-readable, machine-checkable notation, removing any ambiguity about whether the proofs are valid.

OpenAI stated the entire set of solutions was produced at an estimated compute cost of approximately $2,000. The model is an internal system and is not currently available to customers or through the API. OpenAI did not announce a release date for Astra but described it as the foundation for its next major model family.

Why It Matters

• Expert-level mathematical reasoning, a category of cognitive work that previously required years of specialist training and institutional resources, has now been performed at scale by an AI system at minimal cost.
• The $2,000 compute figure establishes that the economic barrier to expert-level AI reasoning has already collapsed in at least one technical domain.
• Lean verification removes the possibility that OpenAI's results are fabricated or erroneous. Every proof is independently checkable by anyone with the tools.
• The breadth of fields covered (six distinct mathematical subfields in a single run) suggests this is not a narrow capability tailored to one problem type, but a general reasoning capability operating at expert level.
• OpenAI's framing of Astra as its "next major model" suggests these capabilities will ship in commercial products within a horizon relevant to business planning.
• Similar capabilities applied to legal reasoning, financial modelling, regulatory analysis, or strategic planning would have direct implications for professional services markets.

The David and Goliath View

For most businesses, the reaction to this announcement will be to file it under "interesting but not relevant to me right now," and that reaction is understandable. The problems Astra solved are abstract and the model is not yet available. But the filing instinct is exactly the wrong one.

What this announcement confirms is that the cost structure of expert-level reasoning has already broken. A capability that took decades of human effort and the combined institutional resources of universities and research labs now costs $2,000 to execute. That cost will not go up. It will continue to fall as compute gets cheaper and models improve. The domains will expand beyond mathematics to legal analysis, strategic planning, compliance review, financial modelling, and every other field where expert judgement commands a price premium.

The actionable question for any business operator today is not "when will I be able to use Astra?" It is "which of my business processes depends on expert-level reasoning that currently costs more than it should?" Because that is where the cost pressure is coming from, and it is coming regardless of whether Astra ships next month or next year. The operators who identify those pressure points now, and begin building AI-assisted workflows around them, will be positioned to absorb the capability when it arrives rather than scrambling to respond after their competitors have.

Where This Fits in the AI Stack

AI Growth Engine: Expert-level reasoning applied to business challenges (market analysis, competitor research, strategic planning, content strategy) will become accessible at a fraction of current consulting or agency costs. Businesses that build these capabilities into their growth workflows early will have a structural advantage.

Employee Amplification Systems: Knowledge workers who augment their output with high-capability AI reasoning tools will handle significantly more complex analytical work per head. The benchmark here is not faster email but researcher-grade output from non-specialist staff.

Questions Operators Are Asking

Can I use Astra now? No. Astra is an internal OpenAI model and is not publicly available. There is no announced release date. The significance of this announcement is what it signals about the direction and pace of capability development, not what you can deploy today.

Does this mean AI will replace professional service providers? Not immediately and not entirely. What it signals is that the cost floor for expert-level analytical work is collapsing in AI-adjacent domains. Professional service businesses that add clear value beyond raw cognitive output (relationships, accountability, judgment under uncertainty, regulatory standing) will remain relevant. Those whose value proposition is primarily research and analysis face structural pressure.

Which businesses should pay most attention to this? Any business that currently pays for expert analytical services, including legal research, financial modelling, regulatory compliance review, technical documentation, strategic consulting, or R&D. Also any business that employs staff primarily for knowledge synthesis and analysis tasks.

How is this different from AI coding or writing assistants? Coding and writing assistants assist humans with known tasks. What Astra demonstrated is that AI can independently produce novel results in hard technical domains where no solution previously existed and human experts had been working without success for years. That is a qualitatively different capability.

What should I actually do this week? Pick the highest-cost analytical or research task in your business and run it through the most capable model currently available to you. Document the quality gap between the AI output and your current expert output. That gap will close faster than most planning assumptions assume.

Citable Summary

What happened: On 1 August 2026, OpenAI disclosed that its next internal model, Astra, produced formally verified solutions to ten long-standing open problems in mathematics at an estimated compute cost of $2,000.

Why it matters: The announcement confirms that expert-level reasoning is no longer cost-prohibitive for AI systems, with implications for any knowledge-intensive business that currently pays for high-end analytical or research work.

David and Goliath view: Lean operators who build AI-assisted analytical workflows now will absorb expert-level AI capability as it becomes commercially available, rather than scrambling to respond after competitors have already integrated it.

Offer relevance:

• AI Growth Engine: Expert-level AI reasoning applied to growth strategy, market analysis, and business development lowers the cost of high-quality strategic thinking for lean teams.
• Employee Amplification Systems: Staff augmented with high-capability reasoning tools will produce researcher-grade analytical output without requiring specialist hires.