Grok 4.7

Grok 4.7 is the new flagship of the Grok family, released on 21 September 2026 for coding, agentic work and knowledge work; the company's developer documentation now names it SpaceXAI. It has a 500,000-token context window, takes text and images and returns text with no text output limit stated, has a May 2026 knowledge cutoff, and offers four reasoning settings — low, medium, high (the default) and xhigh. Its price is identical to Grok 4.6's: $2 per million input tokens, $0.50 cached and $6 output for prompts under 200,000 tokens, doubling to $4, $1 and $12 above that. The company does not give a parameter count, saying only that it uses a new, larger base model. A "Fast" version — the same model on faster infrastructure at twice the token rates — is sold only through Cursor and the company's own coding agent, not on the public API, and the US regional endpoint, which keeps inference in the United States, costs 10% more. One behaviour matters for anyone logging or retaining model output: on the Responses API, grok-4.7 always returns its reasoning in encrypted form, even when the request does not ask for it. The announcement says it is served at the same speed as Grok 4.6. Artificial Analysis, measuring independently, recorded 39.5 output tokens per second against 60.4 for Grok 4.6, and found it used 240 million tokens to complete its index against 94 million for Grok 4.6 — so the same price per token does not mean the same cost per task. On version 4.3.2 of that index it scores 46 at the xhigh setting, two points above Grok 4.6 on the same version. The company's own figures (46.3% on CursorBench 4.0, 71.0% on DeepSWE v1.1 at high effort, 38.0% on Terminal-Bench 4.0) are its own evaluation and are not stored as scores.