Researchers at UC Berkeley have unveiled the Sky-T1-32B-Preview, an open-source reasoning AI model that can be trained for under $450. This model reportedly matches the performance of OpenAI's o1-preview on several benchmarks, achieving scores of 82.4% on Math500 and 86.3% on LiveCodeBench-Easy. The Sky-T1 model is based on a 32 billion parameter architecture and utilizes 17,000 synthetic training data points generated by the Qwen reasoning variant. The training process takes approximately 19 hours, making it a cost-effective alternative for developers looking to create advanced AI systems. The release of Sky-T1 is seen as a potential turning point in the field of reasoning models, drawing comparisons to previous breakthroughs in AI development.