From OpenAI's post https://openai.com/index/navier-stokes-solution/ https://openai.com/index/navier-stokes-solution/ "Since August 28 we have been training a n

From OpenAI's post https://openai.com/index/navier-stokes-solution/ https://openai.com/index/navier-stokes-solution/ "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." "When a further trained version of our internal model became available over the course of the effort" "While unlikely, we cannot rule out that de-identified data derived from their usage of our products helped improve our models "