OpenAI releases 722 math proofs; Mistral launches Le Chonk

OpenAI has published 722 mathematical papers from an unreleased internal model, grouping them into 372 result families that claim to solve or advance major open problems. The release includes a proof of the quasi-Riemann hypothesis and 162 results written in Lean for computer verification, achieved via single prompts averaging three hours of ChatGPT Pro compute. In the open-source sector, Mistral AI launched 'Le Chonk,' a 1T-parameter model that leads non-Chinese open systems on benchmarks, including cybersecurity and legal tasks. These developments highlight a growing trend where complex scientific reasoning is increasingly driven by large-scale AI inference rather than traditional human-led research methods.