DeepSeek-R1 performs reasoning tasks at the same level as OpenAI’s o1 — and is open for researchers to examine.