OpenAI: GeneBench-Pro (openai.com)

OpenAI has announced the launch of GeneBench-Pro, a sophisticated benchmark aimed at evaluating AI models' capabilities in the complex realm of computational biology. Unlike traditional benchmarks, GeneBench-Pro is designed to mimic the judgment-heavy decisions scientists face when analyzing messy, real-world datasets. This new tool encompasses 129 challenging questions that require models to navigate ambiguity, revise assumptions, and select appropriate analytical paths, reflecting the iterative and nuanced nature of scientific research. The introduction of GeneBench-Pro is significant as it seeks to address the current bottleneck in computational biology, where the costs of data generation have dropped but the challenges in analysis remain formidable. The benchmark has shown promising results, particularly with OpenAI's GPT-5.6 Sol model, which achieved a 28.7% pass rate at the highest reasoning level—a significant improvement from previous iterations. This highlights the ongoing advancements in AI's ability to perform complex scientific reasoning, although there's still considerable room for growth. The substantial economic potential of automating such analyses is underscored by the comparison between the costs of human-driven analyses and AI-driven ones, indicating that improved AI models could greatly enhance the efficiency of scientific discovery. GeneBench-Pro sets a precedent for future benchmarks that will not only assess basic knowledge but also the higher-order skills involved in scientific reasoning and judgment.