Grok 4.7 launch — xAI has released Grok 4.7, a model with 2.1 trillion parameters. That parameter count represents a 40% increase over Grok 4.6’s 1.5 trillion parameters and xAI published pricing of $2 per million input tokens and $6 per million output tokens. The update is live now and was released without a waitlist after several timeline revisions that repeatedly moved the date — listed four weeks out, then a few weeks, then three to four weeks, then ten days, and finally described as ‘needs a few more days to cook’.
Grok 4.7 contains 2.1 trillion parameters, which xAI presents as a 40% increase over the 1.5 trillion-parameter count reported for Grok 4.6, reflecting the company’s published model-size figures.
xAI says the model spends longer working through difficult problems, and that it applies this additional processing time to harder prompts relative to Grok 4.6. xAI also says Grok 4.7 double-checks its own answers more often than Grok 4.6 and that it incorporates stronger safety guardrails compared with the prior release.
No additional technical or mechanistic detail about how Grok 4.7 achieves longer problem-solving durations, increased self-checking, or the stronger safety guardrails is provided in the available material.
xAI published pricing for Grok 4.7 as part of its release materials. The published pricing lists costs of $2 per million input tokens and $6 per million output tokens. Those per‑million rates are specified for input tokens and output tokens, respectively. These figures were published alongside the Grok 4.7 model information.
Grok 4.7 recorded a GDPval score of 1695, and Claude Fable 5.1 recorded a GDPval score of 1735. In AA-Briefcase testing, Grok 4.7 scored 1657 while Fable 5.1 scored 1678. The material lists CursorBench 4.0 as pitting Grok 4.7 against GPT-5.6 Sol’s successor GPT-6 Astra, Claude Sonnet 5, and Fable 5.1. These benchmark scores and the CursorBench 4.0 matchup are presented in the available release information.
Grok 4.7 is live now across multiple xAI platforms, including the Grok app, Cursor, Grok Build, and the xAI API, and there is no waitlist. Prior to release, xAI revised the public timeline several times, moving the date from four weeks out to a few weeks, then three to four weeks, then ten days, and finally describing it as ‘needs a few more days to cook’. Elon Musk called Grok 4.7 ‘a strong combination of intelligence, speed & low cost.’ Grok 4.5 launched in July and had third-place scores behind Claude and OpenAI’s models. These points are presented in the release information.
The section above records the release status, timeline revisions, reception, and the prior model’s competitive placement. No additional release details are presented here beyond those reported. The material focuses on the model’s availability and initial reception.
