OpenAI published hundreds of mathematical proofs generated by an unreleased frontier model this week, showcasing progress on mathematical reasoning—a critical benchmark for evaluating advanced AI systems. The release of 719 mathematical manuscripts signals the company's continued push to demonstrate capabilities in complex problem-solving and technical depth.
Meanwhile, open-source AI development gained momentum with two major announcements. Mistral and Reflection AI both launched new open-weight models designed to provide Western alternatives to Chinese AI competitors, reflecting growing industry emphasis on maintaining open-source diversity and preventing AI market consolidation in Asia.
The week also marked another high-profile departure driven by AI safety concerns, as a safety-focused executive announced their resignation. The move continues a pattern of leadership transitions at major AI companies rooted in disagreements over development pace and governance frameworks.