Chinese AI company DeepSeek has launched DeepSeek R1, an open-source large language model that reportedly outperforms OpenAI's models in key benchmarks, particularly in math, coding, and reasoning tasks. R1 is a refined version of its predecessor, DeepSeek R1 Zero, addressing previous issues with readability. Free to use as a web app, R1 offers affordable API access and demonstrates impressive capabilities, suggesting that advanced AI can be both cost-effective and efficient.