“DeepSeek's R1 model replicates what OpenAI has done with its o1 model, achieving similar scale and results in reinforcement learning for reasoning.”