According to Last Week in AI, OpenAI has published 719 mathematical proofs generated by an unreleased frontier model, marking a significant public demonstration of advanced AI mathematical capabilities. The publication did not specify which model produced these proofs or provide details about their mathematical domains.
In related developments, the report notes that Mistral and Reflection AI have launched new open-weight models positioned to compete with Chinese AI offerings. These releases represent continued efforts by Western AI companies to maintain competitiveness in the open-source AI model space, though specific technical details about the models’ capabilities were not provided in the summary.
The report also mentions another safety-related resignation in the AI industry, though details about the individual or organization involved were not included in the available information. These developments come amid ongoing discussions about AI safety and the balance between open and closed AI development approaches.