One of the defining features is its native long-context capability, with a maximum length of 512,000 tokens, 2X OpenAI's GPT-5 family.