Information and Technology Services is pleased to announce the integration of two new models into U-M GPT, the o1 and the o3-mini from OpenAI, which users can now access through the U-M GPT homepage dropdown menu.
This update marks the first inclusion of ‘reasoning models’ to U-M’s AI services, which excel in the ability to perform tasks that require multi-step logical thinking, adaptive reasoning, or more ‘thinking through’ than surface-level analysis. Per OpenAI, these models are optimized for STEM fields and work at a level ‘similar to a PhD student,’ outperforming GPT-4o in tasks for advanced physics, chemistry, and biology. They also demonstrate particular proficiency in generating and debugging complex Python code.
Beyond logically elaborate tasks, both models can serve as versatile tools, enabling users to leverage them as AI-powered assistants for analyzing data, summarizing academic research, generating study guides, translating foreign languages, and more.
Both of these models operate on either a ‘low,’ ‘medium,’ or ‘high’ level of logical reasoning, which refers to its ability to handle tasks with varying degrees of complexity in its reasoning. At present, ITS has set the models to the ‘high’ level, so that users can benefit from the full power of these models and perform tasks that require nuance, depth, and analysis.
This dual integration reflects U-M’s commitment to enhancing and diversifying its AI resources for education, research, and professional development. As always, these and all other models which U-M GPT supports remain available at no cost to the U-M community. For more details on AI at U-M, please visit our AI services page.
