xAI Ships Grok 4.6 on the Same 1.5T Foundation as 4.5 — With No Model Card, No Benchmarks, and a 2.1T Grok 4.7 Already Teed Up

Evolink (secondary; no xAI model card published)·low signalXBlueskyLinkedInCopy link

Grok 4.6 launched Aug 7 reusing the 1.5-trillion-parameter V9 foundation from Grok 4.5, with gains attributed entirely to improved supervised fine-tuning and RL rather than scale. As of launch xAI has published no benchmarks, model card, or scores, so every number circulating is speculation until independent Arena results land next week; for reference Grok 4.5 (high) sits at 54 on the Artificial Analysis Intelligence Index, fourth behind Claude Fable 5, GPT-5.5 and Claude Opus 4.8. A larger 2.1T-parameter Grok 4.7 is expected within weeks. Flagged low because the primary source is secondary reporting and xAI has published nothing verifiable.