Open AI cost reset

The 68-hour pretraining claim proves broadly credible and accelerates the view that capable small models can be trained with far less capital than assumed. Open-weight competitors, enterprise AI teams, and application developers use the Thinky Machines result to reduce dependence on frontier-scale training budgets.

Each figure is the modeled move of the S&P 500 if the outcome occurs. For information only. Not investment advice.