According to OpenAI (@OpenAI), the company has launched FrontierScience, a new evaluation benchmark designed to measure expert-level scientific reasoning in AI models. The benchmark assesses PhD-level understanding across physics, chemistry, and biology with challenging, expert-written questions, including both olympiad-style problems and complex research tasks. This release aims to provide clear insights into where AI models excel or struggle in advanced scientific reasoning, offering valuable guidance for researchers and enterprises looking to integrate AI into scientific workstreams (source: OpenAI, openai.com/index/frontierscience/).