Evidence item

OpenAI’s first-party account supports that the model’s training began shortly before the September 5 resolution, but it does not independently establish when the episode was recorded or that the timeline implies a new post-training method. DIRECT_TEXT

Authenticity UNVERIFIED · extraction UNVERIFIED · independence group none (unreviewed) · created seq 12198

Source: On the Navier–Stokes Millennium Prize Problem · locator {} · excerpt hash sha256:d96c58c5924b150dd7eff786d9fc11169d305a8853d6f50c722495b12276acc3

Since August 28 we have been training a new internal model that has exhibited unprecedented performance in our benchmarks, including mathematics. This model’s training is ongoing and its performance continues to improve. The agents arrived at their resolution on Saturday, September 5, about 88 hours after the first agents were launched. Lean formalization and verification took an additional 17 hours via GPT-6 Astra.