Grok 4.7

xAI's current flagship: 'SpaceXAI's frontier model for coding, agentic tasks, and knowledge work' per the official model page, and the models index directs general use to it ('For everything else, including code, use Grok 4.7'). Launched on the API 2026-09-21, dated by xAI's own release notes. Modality 'text, image -> text' per the official model page. No alias is published for this id - the Details panel lists none, and the models index carries only grok-4.5-latest and grok-4.3-latest. TIERED pricing: figures shown are the base tier for prompts <=200K tokens; prompts >200K are billed 2x ($4/M input, $1/M cached, $12/M output). Knowledge cut-off May 2026, stated in prose on the models index with no day given, so the field carries the first of the month as a convention and NOT as a published date. Max output tokens null as a SOURCED ABSENCE OF A LIMIT, not a missing value: the 2026-09-21 release note states 'no text output limit'. Reasoning efforts low/medium/high/xhigh, default high. Batch API NOT supported, same as grok-4.6 and unlike grok-4.3 and grok-4.5. On the Responses API it always returns reasoning.encrypted_content even when include does not list it. Rates above are the us-east-1 cluster, which is what the model page shows by default; the pricing page's own per-cluster payload prices us-west-2 identically and us-central-1 10% higher ($2.20/$0.55/$6.60), exactly as it already does for grok-4.6. NOT CARRIED: 'Grok 4.7 Fast', which the release note describes as the same model at twice the token rates and states is available only through Cursor and Grok Build, not on the public xAI API.

Source: xAI official documentation ↗

Compare Grok 4.7 head-to-head

Grok 4.7 — questions & answers