OpenAI says an internal model solved a major math problem

OpenAI says an unreleased internal model helped produce a proof for NavierStokes, one of the seven Millennium Prize problems with a $1 million reward. The company said the work was done with about 10,000 AI agents over 88 hours and described the model as more capable than GPT6 Astra. The claim has drawn attention because mathematicians Tristan Buckmaster and Levent Alpöge had been pursuing a similar approach for about a year. Buckmaster questioned whether OpenAI may have benefited from drafts or related work, while OpenAI said it had not seen their work and did not access specific user data. The newsletter also noted that OpenAI released ChatGPT Images 2.5, an updated image model that generates faster and improves editing. The update adds tools such as sketching, templates, comments, and shared prompts, and the company said the model is now better at changing only the parts of an image users ask to edit.