On September 21, SpaceX AI officially launched the Grok 4.7 model. The official selling points are strong: the reasoning speed is twice that of comparable models, while the price is only half of the competitor's. Specifically, the cost for input tokens is $2 per million, and $6 per million for output tokens. This directly raises the bar for cost-effectiveness by making advances on both speed and cost.