In a meeting with Meta employees on Thursday, Mr. Zuckerberg also doubled down on recent changes to the company’s online speech policies and ending its diversity initiatives.
China AI startup DeepSeek just released its R-1 model that compares favorably with OpenAI's o1 reasoning model. DeepSeek claims to have trained R1 at a fraction of the cost of o1 and Meta's Llama 3.1.