DEEPSEEK CRUSHES OPENAI O1 WITH AN MIT-LICENSED MODEL—DEVELOPERS ARE LOSING IT

Summary from the TLDR newsletter

Chinese AI research lab DeepSeek has unveiled its latest reasoning models, DeepSeek-R1 and DeepSeek-R1-Zero. DeepSeek-R1 is fully open-source and distributed under the MIT license. The lab has released six distilled models, ranging from 32 billion to 70 billion parameters. They are designed to address tasks in math, code generation, and reasoning.