New AI Model o3-mini Introduced by Stanford, UW, and Others, 63% Cheaper and Faster Than o1-mini
Researchers from Stanford University, the University of Washington, the Allen Institute for AI, and Contextual AI have introduced a new approach to test-time scaling for large language models (LLMs). This method aims to enhance reasoning capabilities and improve performance, particularly in comparison to OpenAI's o1 model. The new model, referred to as o3-mini, is reported to be faster, smarter, and 63% cheaper than the previous o1-mini, and 93% cheaper than the original o1 model. This development is part of a broader shift in AI towards models that utilize slow thinking, step-by-step reasoning, and self-correction, which are crucial for advancing AI research and applications.
Sources
Brian Roemmele
The 1000 Questions That Change Small AI Models That Beat Large Models — The new paper titled "s1: simple Test-time Scaling approach to exceed OpenAI’s o1-preview performance" presents a significant advancement in the field of large language models (LLMs) and their reasoning… https://t.co/ZJUKx9pPqS
TuringPost Since @OpenAI's o1 model release, everything turned upside down with the shift to models using: - slow "thinking" - step-by-step reasoning (Chain-of-Thought) - self-correction These aspects link to a fascinating test-time compute topic. We've summarized what's going on in this… https://t.co/jWFVZFDXhw
Vlad Ruso PhDs1: A Simple Yet Powerful Test-Time Scaling Approach for LLMs #LanguageModels #TestTimeScaling #AIResearch #MachineLearning #InnovationInAI https://t.co/m3EmedYDIV https://t.co/YJp0mfCxVD
Additional media
