developing · Story package
OpenAI reports formal-mathematics results from an internal model
The company says an internal model produced ten new results and is releasing manuscripts, reasoning walkthroughs, and Lean certificates for outside examination.
Overview
OpenAI says an internal version of its next major model produced ten new results on long-standing mathematics and theoretical-computer-science problems at an estimated cost of roughly $2,000 using GPT-5.6 Sol API rates. The company says manuscripts, reasoning walkthroughs, and formal Lean certificates are being released for external examination, so the claims remain pending that review.
Why it matters
Externally checkable manuscripts and formal certificates could provide stronger evidence than benchmark-only claims if the results withstand review.
The number, novelty, cost, and model-capability claims currently come from OpenAI and have not yet been independently validated.
Key facts
OpenAI says an internal model produced ten new results on long-standing mathematics and theoretical-computer-science problems.
OpenAI says it is releasing manuscripts, reasoning walkthroughs, and formal Lean certificates for external examination.
The company estimates the work at roughly $2,000 using GPT-5.6 Sol API rates.
Latest update
No public update is available.
Full timeline
No public timeline entries are available.
Sources
Open questions
- Will independent experts confirm the novelty and correctness of all ten results?
- What role did humans, tools, search, and formal verification play in producing the results?
- Does the stated cost include the complete generation and verification workflow?
Related stories
No related stories are listed.
Roundup appearances
No roundup appearances are available.