DeepSeek, which in late November unveiled DeepSeek-R1, an answer to OpenAI’s o1 “reasoning” model, is a curious organization.

The new model shows open-source closing in on closed-source models, suggesting reduced chances of one big AI player ruling the game.