Model Overview & Capabilities
Grok 4.7 is xAI’s updated frontier intelligence model released on September 30, 2026, ranking #12 on our global leaderboard with a composite index score of 43.6. Superseding the earlier Grok 4.6 generation (44.0 index), Grok 4.7 features a refined chain-of-thought reasoning harness, achieving verified scores of 44.0% on SWE-bench Verified, 43.5% on Terminal-Bench 2.0, 44.0% on Aider Polyglot, and 86.5% on GPQA Diamond graduate scientific evaluations.
Billed at $2.00 per million input tokens and $8.00 per million output tokens, Grok 4.7 produces a blended cost of $3.50 per million tokens under standard 3:1 input:output weighting. Prompt caching is supported at $0.20 per million tokens, achieving a respectable value efficiency of $0.0803 per index point. The model streams tokens at 105 tokens per second with an average time to first token of 1.50 seconds across production developer queries.
With a 500,000-token context window and a 32,768-token output limit, Grok 4.7 provides strong utility for complex mathematical algorithm optimization, real-time API contract verification, automated test synthesis, and autonomous software issue triage in continuous integration developer workflows. Engineering organizations utilize Grok 4.7 for demanding code analysis tasks that demand high factual rigor and reliable mathematical verification.
