Overview
- SpaceXAI has made Grok 4.7 available through Cursor, Grok Build and a public Grok API and says the model was trained natively on its Grok Bot harness to improve dialogue, document and presentation generation.
- The company reports benchmark gains over Grok 4.6 on professional tests such as GDPval and AA Briefcase and says performance now matches other leading models for knowledge‑work tasks.
- SpaceXAI published tokenized pricing of $2 per million input tokens and $6 per million output tokens and an optional higher‑speed tier that the company says costs roughly double.
- SpaceXAI announced a new safety regime and shared test results for biosecurity and dual‑use prompts, and it is offering invite‑only access to selected cybersecurity partners for red‑teaming and defensive work.
- Coverage notes a conflict in SpaceXAI’s own statements about speed and cost—some promotional claims say Grok 4.7 is much faster and cheaper than peers while another company line says runtime and pricing match Grok 4.6—and the published benchmarks and tests remain company‑reported and not yet independently verified.