News Summary

Researchers from Stanford University and the University of Washington have developed an AI reasoning model named ‘s1’ for under $50, challenging leading models from OpenAI and DeepSeek. The model was trained using a method called ‘distillation,’ leveraging the outputs of Google’s Gemini 2.0 Flash Thinking Experimental model. Remarkably, training s1 took less than 30 minutes on 16 Nvidia H100 GPUs, costing about $20 in computing power. The model is publicly accessible on GitHub, advocating for transparency and collaboration in AI development. This achievement raises questions about AI accessibility and the diminishing competitive edge of well-funded AI labs, as high-performance models become replicable with minimal investment.