DeepSeek-R1 Open-Sourced, Inference Cost Only 3% of OpenAI's

Chinese AI company DeepSeek released and fully open-sourced its R1 reasoning model, which matches OpenAI's o1 on tasks like math reasoning and code generation, but with training costs less than 3% of the latter. This news shocked the global AI industry, caused NVIDIA's stock to plummet, and reignited global attention on China's AI capabilities.