SpaceXAI has released Grok 4.7, which secured the fourth position among frontier AI labs according to Artificial Analysis. The model achieved a score of 46 on the Intelligence Index, marking a 2-point increase over its predecessor, Grok 4.6. It currently ranks behind Claude Fable 5.1, GPT-6 Astra, and Claude Opus 5.
In professional benchmarking, Grok 4.7 reached 1657 Elo on AA-Briefcase and 1695 Elo on GDPval-AA, showing improvements of 111 and 90 points respectively. The model also scored 56 on the Coding Agent Index, surpassing GPT-5.6 Sol. However, performance metrics on AA-LCR and AutomationBench-AA tests saw a decline compared to previous versions.
The pricing structure remains unchanged at $2 per million input tokens and $6 per million output tokens, with a 500 000-token context window. Despite stable rates, the cost per task is higher because Grok 4.7 utilizes approximately 81 000 output tokens per index task, compared to 36 000 for Grok 4.6.
The increased token consumption impacts a division that previously reported a loss of $1,26 billion. Elon Musk stated that the model utilizes a 2,1-trillion-parameter base and incorporates SpaceX company data. Development continues, with Grok 4.8 training nearing completion and expectations set for Grok 5 to reach artificial general intelligence.
