Elon Musk Said Grok 4.7 Would Exceed Every Model. Did It Deliver?

SpaceX's AI division, SpaceXAI, has officially released Grok 4.7, pushing the company into the ranks of the top four frontier AI laboratories globally. According to independent benchmark testing by Artificial Analysis, Grok 4.7 achieved a score of 46 on the Intelligence Index, marking a 2-point increase over Grok 4.6, though it still trails Claude Fable 5.1, GPT-6 Astra, and Claude Opus 5. The model, built on a 2.1-trillion-parameter base and trained partly on proprietary SpaceX company data, demonstrated notable improvements in professional tasks and coding capabilities, scoring 56 on the Coding Agent Index to overtake GPT-5.6 Sol. Despite performance improvements, the gains highlight mounting compute inefficiencies. While headline pricing remains unchanged at $2 per million input tokens and $6 per million output tokens, Grok 4.7 generates roughly 81,000 output tokens per Index task—significantly higher than Grok 4.6's 36,000 and GPT-6 Astra's 27,000—driving up the effective cost per query. These compute demands arrive as SpaceXAI faces scrutiny over a reported $1.26 billion divisional loss, even as Elon Musk targets the upcoming Grok 4.8 and Grok 5 models for further benchmark climbs toward artificial general intelligence.