xAI released Grok 4.7 starting at 2 dollars per million input tokens, below the models it benchmarks against

Sources: primary, discussion
Summary: xAI prices Grok 4.7 starting at 2 dollars per million input tokens and 6 per million output, states it also serves a fast variant with twice the output speed at twice that price, and states the model is available in Cursor, the Grok API, and third-party harnesses and routers. The vendor's own comparison table covers Grok 4.7, Grok 4.6, GPT-5.6 Sol Max, and Fable 5.1 Max across seven benchmarks, and Fable 5.1 Max scores higher on four of them: CursorBench 4.0 at 51.8 against 46.3, Terminal-Bench 4.0 at 57.9 against 38.0, HealthBench Professional at 62.1 against 56.7, and AA Briefcase v1.1 at 1,678 against 1,657, while GPT-5.6 Sol leads DeepSWE v1.1 at 72.7 against a 71.0 high-effort score and Grok 4.7 leads EEBench and the Harvey legal benchmark. Every figure is vendor-reported and none is independently reproduced, and the main table omits GPT-6 Astra, which appears only in a later chart on the same page.
Why it matters: The release competes on price rather than on scores, at one fifth the input list price and roughly one eighth the output list price of Fable 5.1 Max, a model the vendor's own table places ahead of Grok 4.7 on four of the seven listed benchmarks.
Follow-up: Whether independent harnesses reproduce the vendor-reported benchmark figures.