Grok 4.6 missed its August 7 target date. As of August 10, xAI's own release notes still show none.

Elon Musk said on July 28 that [xAI](#/company/xai)'s next model, Grok 4.6, would release “around August 7,” describing it in a post on X as the 1.5-trillion-parameter model with “significantly improved SFT & RL,” with a larger, 2.1-trillion-parameter Grok 4.7 following “weeks” later.

As of August 10, three days past that target, xAI's own developer release notes show no such model. The newest text model listed is July's Grok 4.5, priced at $2 per million input tokens and $6 per million output tokens; there is no model card, identifier, pricing, or availability entry for Grok 4.6, and none for Grok 4.7 either. xAI's release notes are the operative record here — a shipped model needs an endpoint developers can actually call, not just a name.

Grok 4.5 — the model xAI's release notes still list as newest — carries an independent [Artificial Analysis](#/scoreboard) Intelligence Index score of 54 at that same $2/$6 rate, a real, measured position in the current field rather than a promise. Grok 4.6 and 4.7 have no equivalent yet, because neither has an API endpoint, a model card, or weights an independent evaluator could actually test.

Neither the 1.5-trillion-parameter spec Musk gave for Grok 4.6 nor the 2.1-trillion-parameter figure he gave for Grok 4.7 has been confirmed by xAI itself — both numbers come only from Musk's own post. Until either model appears in the release notes with a model card and published pricing, both remain announced but unshipped, and this publication's Scoreboard will carry them as unranked once they do, the same way it currently handles every other released-but-unmeasured launch.